IJCSIS Computer Science Research

Report as inappropriate Your report has been received

Description

International Journal of Computer Science and Information Security (IJCSIS) is a peer-reviewed monthly journal devoted to research and applications of computer science and security. The journal papers follow IEEE format guidelines and publication standards. IJCSIS editorial board consists of several internationally recognized experts and guest editors. Wide circulation is assured because libraries and individuals, worldwide, subscribe and reference to IJCSIS. The Journal has grown rapidly to its currently level of about 1,000 articles published and indexed. Other field coverage includes: security infrastructures, network security: Internet security, content protection, cryptography, steganography and formal methods in information security; multimedia systems, software, information systems, intelligent systems, web services, data mining, wireless communication, networking and technologies, innovation technology and management. (See monthly Call for Papers) IJCSIS is published using an open access publication model, meaning that all interested readers will be able to freely access the journal online without the need for a subscription. On behalf of the Editorial Board and the IJCSIS members, we would like to express our gratitude to all authors and reviewers for their hard and high-quality work. Available at http://sites.google.com/site/ijcsis/ IJCSIS Vol. 9, No. 3, March 2011 Edition ISSN 1947-5500 � IJCSIS, USA.

Shared by: ijcsis

Stats

views:: 2971
posted:: 4/12/2011
language:: English
pages:: 318

Public Domain

Document Sample

IJCSIS Vol. 9 No. 3, March 2011
ISSN 1947-5500

International Journal of
Computer Science
& Information Security

IJCSIS editorial board consists of several internationally recognized experts and guest editors.
Wide circulation is assured because libraries and individuals, worldwide, subscribe and reference
to IJCSIS. The Journal has grown rapidly to its currently level of about 1,000 articles published
and indexed.

Other field coverage includes: security infrastructures, network security: Internet security,
content protection, cryptography, steganography and formal methods in information security;
multimedia systems, software, information systems, intelligent systems, web services, data
mining, wireless communication, networking and technologies, innovation technology and
management. (See monthly Call for Papers)

IJCSIS is published using an open access publication model, meaning that all interested readers
will be able to freely access the journal online without the need for a subscription.

On behalf of the Editorial Board and the IJCSIS members, we would like to express our gratitude
to all authors and reviewers for their hard and high-quality work.

Available at http://sites.google.com/site/ijcsis/
IJCSIS Vol. 9, No. 3, March 2011 Edition
ISSN 1947-5500 © IJCSIS, USA.

Abstracts Indexed by (among others):
IJCSIS EDITORIAL BOARD

Dr. M. Emre Celebi,
Assistant Professor, Department of Computer Science, Louisiana State University
in Shreveport, USA

Dr. Yong Li
School of Electronic and Information Engineering, Beijing Jiaotong University,
P. R. China

Prof. Hamid Reza Naji
Department of Computer Enigneering, Shahid Beheshti University, Tehran, Iran

Dr. Sanjay Jasola
Professor and Dean, School of Information and Communication Technology,
Gautam Buddha University

Dr Riktesh Srivastava
Assistant Professor, Information Systems, Skyline University College, University
City of Sharjah, Sharjah, PO 1797, UAE

Dr. Siddhivinayak Kulkarni
University of Ballarat, Ballarat, Victoria, Australia

Professor (Dr) Mokhtar Beldjehem
Sainte-Anne University, Halifax, NS, Canada

Dr. Alex Pappachen James, (Research Fellow)
Queensland Micro-nanotechnology center, Griffith University, Australia

Dr. T.C. Manjunath,
ATRIA Institute of Tech, India.
TABLE OF CONTENTS

1. Paper 28021153: Addressing Vulnerability of Mobile Computing: A Managerial Perspective (pp.
1-5)

Arben Asllani and Amjad Ali
Center for Security Studies, University of Maryland University College, Adelphi, Maryland, USA

2. Paper 28021166: Reduction of PAPR for OFDM Downlink and IFDMA Uplink Wireless
Transmissions (pp. 6-13)

Bader Hamad Alhasson , Department of Electrical and Computer Engineering, University of Denver,
Denver, USA
Mohammad A. Matin, Department of Electrical and Computer Engineering, University of Denver, Denver,
USA

3. Paper 28021123: Evolution Prediction of the Aortic Diameter Based on the Thrombus Signal from
MR Images on Small Abdominal Aortic Aneurysms (pp. 14-19)

A. Suhendra, C.M. Karyati, A.Muslim, A.B. Mutiara
Faculty of Computer Science and Information Technology, Gunadarma University, Jl. Margonda Raya
No.100, Depok 16424, Indonesia

4. Paper 28021145: Empirical Evaluation of the Shaped Variable Bit Rate Algorithm for Video
Transmission (pp. 20-29)

A. Suki M. Arif, Suhaidi Hassan, Osman Ghazali, Mohammed M. Kadhum
InterNetWorks Research Group, UUM College of Arts and Sciences Universiti Utara Malaysia, 06010
UUM Sintok, Kedah, Malaysia

5. Paper 28021146: An Efficient Self-Organized Authentication and Key Management Scheme for
Distributed Multihop Relay-Based IEEE 802.16 Networks (pp. 30-38)

Adnan Shahid Khan, Norsheila Fisal, Sharifah Kamilah, Sharifah Hafizah, Mazlina Esa, Zurkarmawan
Abu Bakar
UTM-MIMOS Center of Excellence in Telecommunication Technology, Faculty of Electrical Engineering,
Universiti Teknologi Malaysia 81310 Skudai, Johor, Malaysia
M. Abbas, Wireless Communication Cluster, MIMOS Berhad, Technology Park Malaysia, 57000 Kuala
Lumpur, Malaysia

6. Paper 28021157: A Digital Image Encryption Algorithm Based On Chaotic Logistic Maps Using A
Fuzzy Controller (pp. 39-44)

Mouad HAMRI & Jilali Mikram, Mathematics and computer science department, Science University of
Rabat-Agdal, 4 Avenue Ibn Battouta Rabat Morocco
Fouad Zinoun, Economical sciences and management department, University of Meknes Morocco

7. Paper 28021159: Performance Analysis of Connection Admission Control Scheme in IEEE 802.16
OFDMA Networks (pp. 45-51)

Abdelali El Bouchti, Said El Kafhali and Abdelkrim Haqiq
Computer, Networks, Mobility And Modeling Laboratory, e- NGN Research Group, Africa And Middle
East, FST, Hassan 1st University, Settat, Morocco
8. Paper 23021105: Enhancement and Minutiae Extraction Of Touchless Fingerprint Image Using
Gabor And Pyramidal Method (pp. 52-57)

A. John Christopher, Associate Professor, Department of Computer Science, S.T. Hindu College, Nagercoil,
Dr. T. Jebarajan, Principal, V.V. College of Engineering, Tisayanvilai

9. Paper 23021108: Automatic Parsing For Arabic Sentences (pp. 58-63)

Zainab Ali Khalaf, School of computer science, (USM), Penang, Malaysia
Dr. Tan Tien Ping, School of computer science, Universiti Sains Malaysia (USM), Penang, Malaysia

10. Paper 28021129: Amelioration of Walsh-Hadamard Texture Patterns based Image Retrieval
using HSV Color Space (pp. 64-69)

Dr. H.B.Kekre, Sudeep D. Thepade, Varun K. Banura
Computer Engineering Department, MPSTME, SVKM’s NMIMS (Deemed-to-be University), Mumbai,
India

11. Paper 27021119: Analysis and Comparison of Medical Image Fusion Techniques: Wavelet based
Transform and Contourlet based Transform (pp. 70-75)

C G Ravichandran, RVS College of Engg. & Tech, Dindigul
R. Rubesh Selvakumar, Research Scholar, Anna University of Technology, Tricirappalli
S. Goutham, Surya Engineering College, Erode

12. Paper 28021135: Performance Comparison of Texture Pattern Based Image Retrieval Methods
using Walsh, Haar and Kekre Transforms with Assorted Thresholding Methods (pp. 76-83)

Dr. H. B. Kekre, Sudeep D. Thepade, Varun K. Banura
Computer Engineering Department, MPSTME, SVKM’s NMIMS (Deemed-to-be University), Mumbai,
India

13. Paper 28021139: A Generic Rule-Based Agent for Monitoring Temporal Data Processing (pp. 84-
89)

S. Laban, International Data Centre (IDC), Comprehensive Nuclear Test-Ban Treaty Organization (CTBT),
Vienna, Austria
A.I. El-Desouky, Computer and Systems Department, Faculty of Engineering, Mansoura University,
Mansoura, Egypt
A. S. ElHefnawy, Information Technology, Department, Faculty of Computer, & Information, Mansoura
University, Mansoura, Egypt

14. Paper 28021141: A New Approach for Model based Gait Signature Extraction (pp. 90-94)

Mohamed Rafi, Dept. of CS&E, HMS Institute of Tech., Tumkur, Karnataka, India
Shanawaz Ahmded J, College of Computer Science, King Khalid University, Abha, Kingdom of Saudi Arabia
Md. Ekramul Hamid, College of Computer Science, King Khalid University, Abha, Kingdom of Saudi Arabia
R.S.D Wahidabanu, Dept. of E&C, Government college of Engg Salem, Tamil Nadu, India.

15. Paper 28021142: Mining Fuzzy Cyclic Patterns (pp. 95-99)

A. Mazarbhuiya, M. A. Khaleel, P. R. Khan
Department of Computer Science, College of Computer Science, King Khalid University, Abha, Kingdom of
Saudi Arabia
16. Paper 28021144: Robust Color Image Watermarking Using Nonsubsampled Contourlet
Transform (pp. 100-111)

C. Venkata Narasimhulu , Professor ,Dept of ECE,Hasvita Institute of Engg & Tech,Hyderabad,INDIA
K. Satya Prasad, Professor, Dept of ECE, JNTU Kakinada,India

17. Paper 28021150: Parallel Implementation of Compressed Sensing Algorithm on CUDA- GPU (pp.
112-119)

Kuldeep Yadav & Ankush Mittal, Computer Science and Engineering, College of Engineering
Roorkee , Roorke-247667, India
M. A. Ansar & Avi Srivastava, Galgotia College of Engineering, Gr. Noida, India

18. Paper 28021151: Fuzzy HRRN CPU Scheduling Algorithm (pp. 120-124)

Bashir Alam, M.N. Doja, Department of Computer Engineering, Jamia Millia Islamia, New Delhi, India
R. Biswas, Department of Computer Science and Engineering, Manav Rachna University, Faridabad, India
M. Alam, Department of computer Science, Jamia millia Islamia,New Delhi India

19. Paper 28021152: Experiences and Comparison Study of EPC & UML For Business Process & IS
Modeling (pp. 125-133)

Md. Rashedul Islam School of Business and Informatics Högskolan i Borås Borås, Sweden
Md. Rofiqul Islam School of Business and Informatics Högskolan i Borås Borås, Sweden
Md. Shariful Alam School of Business and Informatics Högskolan i Borås Borås, Sweden
Md. Shafiul Azam Dept. of Computer Science and Engineering Science and Technology University, Pabna
Pabna, Bangladesh

20. Paper 28021155: Facial Tracking Using Radial Basis Function (pp. 134-138)

P. Mayilvahanan, Research scholar, Dept. of MCA, Vel's University, Pallavaram, Chennai, India
Dr. S. Purushothaman, Principal, Sun College of Engineering & Technology, Kanyakumari – 629902,
Tamil Nadu, India
Dr. A. Jothi, Dean, School of Computing Sciences, Vels University, Pallavaram, Chennai, India

21. Paper 28021156: Performance Comparison of Speaker Identification using circular DFT and
WHT Sectors (pp. 139-143)

Dr. H B Kekre, Vaishali Kulkarni, Indraneal Balasubramanian, Abhimanyu Gehlot, Rasik Srinath
MPSTME, NMIMS University.

22. Paper 28021158: Reliability and Security in MDRTS: A Combine Colossal Expression (pp. 144-
153)

Gyanendra Kumar Gupta, Computer Science & Engineering Department, Kanpur Institute of Technology,
Kanpurr, UP, India, 208 001
A. K Sharma, Computer Sc. & Engg. Deptt, M.M.M. Engineering College, Gorakhpur, UP, India, 273010
Vishnu Swaroop, Computer Science & Engineering Department, M.M.M. Engineering College, Gorakhpur,
UP, India, 273010
23. Paper 28021161: Implementation in Java of a Cryptosystem using a Dynamic Huffman Coding
and Encryption Methods (pp. 154-159)

Eugène C. Ezin
Institut de Mathématiques et de Sciences Physiques, Unité de Recherche en Informatique et Sciences
Appliquées, Université d’Abomey-Calavi, République du Bénin

24. Paper 28021164: Towards Generating a Rulebase to Provide Feedback at Design Level for
Improving Early Software Design (pp. 160-164)

B. Bharathi, Research Scholar, Sathyabama University, Chennai-119
G. Kulanthaivel , Assistant Professor, NITTTR, Chennai

25. Paper 28021165: Performance Comparison of TCP Variants in Mobile Ad- Hoc Networks (pp.
165-170)

Mandakini Tayade, School of Information Technology, Rajiv Gandhi Prodyogiki Vishwavidyalaya, Bhopal
(M.P.) India
Sanjeev Sharma, Head of School of Information Technology, Rajiv Gandhi Prodyogiki Vishwavidyalaya,
Bhopal (M.P.) India

26. Paper 28021168: Analysis on Robust Adaptive Beamformers (pp. 171-178)

T. S. Jeyali Laseetha, Professor, Department of Electronics and Communication Engineering, Anna
University of Technology, Tirunelveli, Tamil Nadu, India
Dr. (Mrs) R.Sukanesh, Professor, Department of Electronics and Communication Engineering, Madurai,
Tamil Nadu, India

27. Paper 28021169: A Review On Distance Measurement And Localization In Wireless Sensor
Network (pp. 179-184)

Kavindra Kumar Ahirwar, School Of Information Technology, Rajiv Gandhi Technical University, Bhopal
(MP), India
Dr. Sanjeev Sharma (Head of department), School Of Information Technology, Rajiv Gandhi Technical
University, Bhopal (MP), India

28. Paper 27021117: An Improved Visual Cryptography Scheme for Secret Hiding (pp. 185-197)
Full Text: PDF

G. Prasanna Lakshmi, Computer Science, IBSAR, Karjat, India
Dr. J.A.Chandulal, Professor and HOD, IBSAR, Computer Science, India
Dr. KTV Reddy, Professor & Principal, Electronics & Telecommunications Dept., Computer Science, India

29. Paper 17021103: Adaptive MIMO-OFDM Scheme with Reduced Computational Complexity and
Improved Capacity (pp. 198-205)

L. C. Siddanna Gowd, A. R. Ranjini and M. Kanthimathi,
Faculty of ECE Dept, SriSairam Engineering College, Chennai, T.N., India
30. Paper 24021111: An Efficient Fair Queuing Model for Data Communication Networks (pp. 206-
216)

M. A. Mabayoje 1* , S. O. Olabiyisi 2, A.O. Ameen 1, R. Muhammed 1 , O.C. Abikoye 1.
1
Department of Computer Science, Faculty of Communication and Information Sciences, University of
Ilorin, PMB1515, Ilorin, Kwara-Nigeria.
2
Department of Computer Science and Engineering, Ladoke Akintola University of Technology,
Ogbomosho, Oyo-Nigeria.

31. Paper 24021112: Implementation of Audio Wave Steganography By Replacing 4th Bit LSB of
Audio Wave File (pp. 217-219)

Mr. Vijay B. Gadicha, Department of Computer Science & Engg, P.R.Patil College of Engg & Tech,
Amravati (MH),India.
Mr. Ajay. B. Gadicha, Department of Computer Science & Engg, P.R.Pote(Patil) College of Engg & Tech,
Amravati (MH),India.

32. Paper 24021114: Modeling of Aluminium – Flyash Particulate Metal Matrix Composites using
Fuzzy Logic (pp. 220-225)

R. Elangovan, Research Scholar, Department of Mechanical Engineering, Vinayaka Missions University,
Salem, India-636 308
Dr. S. Purushothaman, Principal, Sun College of Engineering and Technology, Sun Nagar, Erachakulum,
Kanyakumari District – 629902, India

33. Paper 27021116: Compression Techniques and Water Marking of Digital Image using Wavelet
Transform and SPIHT Coding (pp. 226-258)

G. Prasanna Lakshmi, Computer Science, IBSAR, Karjat, India
Dr. J. A. Chandulal, Professor and HOD, IBSAR, Computer Science, India
Dr. KTV Reddy, Professor & Principal, Electronics & Telecommunications Dept., Computer Science, India

34. Paper 28021171: An analytical survey on Network Security Enhancement Services (pp. 259-262)

Deshraj Ahirwar , PG Scholar, CSE, SATI, Vidisha
Manish K. Ahirwar , CSE, UIT, RGPV
Piyush K. Shukla, CSE, UIT, RGPV
Pankaj Richharia, CSE, BITS, Bhopal

35. Paper 28021199: An Empirical Study of Software Project Management among Some Selected
Software Houses in Nigeria (pp. 263-271)

Olalekan Akinola, Funmilayo Ajao, Opeoluwa B. Akinkunmi
Computer Science Department, University of Ibadan, Ibadan, Nigeria

36. Paper 30011123: New Codes for Spectral Amplitude Coding Optical CDMA Systems (pp. 272-279)

Hassan Yousif Ahmed, Communication & Networking Engineering Department, Computer Science
College, King Khalid University, Abha, Kingdom of Saudi Arabia
Elmaleeh, M. A, Electronics Engineering Dept, Faculty of Engineering and Technology, University of
Gezira, Wad Madni, Sudan
Hilal Adnan Fadhil, School of Computer and Communication Engineering, Universiti Malaysia Perlies,
Malaysia
S.A. Aljunid, School of Computer and Communication Engineering, Universiti Malaysia Perlies, Malaysia
37. Paper 31011135: Image Retrieval with Image Tile Energy Averaging using Assorted Color Spaces
(pp. 280-286)

Dr. H.B.Kekre, Sudeep D. Thepade, Varun Lodha, Pooja Luthra, Ajoy Joseph, Chitrangada Nemani
Information Technology Department, MPSTME, SVKM‟s NMIMS (Deemed-to-be University), Mumbai,
India

38. Paper 27021118: Improvement of Distributed Virtual Environment (DVE) performance (pp. 287-
295)

Olfat I. EL-Mahi, Computer Graphics department, IRI institute- MuCSAT, Borg EL-Arabe, Egypt
Hanan Ali, Computer Graphics department, IRI institute- MuCSAT, Borg EL-Arab, Egypt
Walaa M. Sheta, Computer Graphics department, IRI institute- MuCSAT, Borg EL-Arab, Egypt
Salwa Nassar, Electronic Research Institute, Cairo, Egypt
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9 No. 3, March 2011

Addressing Vulnerability of Mobile Computing
A Managerial Perspective

Arben Asllani and Amjad Ali
Center for Security Studies
University of Maryland University College
Adelphi, Maryland, USA

Abstract— Popularity of mobile computing in organizations has approximately 0.6 percent in stock price when a vulnerability is
risen significantly over the past few years. Notebooks and laptop reported and the impact is more severe when the vulnerability
computers provide the necessary computing power and mobility flaws are not addressed in advance [2]. However, while most
for executives, managers, and other professionals. Such organizations consider vulnerability management critical to
advantages come with a price for the security of the their operations, fewer than 25 percent have vulnerability as an
organizational networks: increased vulnerability. The paper integrated part of their operations [3]. This paper offers a
discusses three types of mobile computing vulnerability: physical, managerial framework to address the issues of information
system, and network access vulnerability. Using a managerial systems vulnerabilities with a special focus on laptop
approach, the paper offers a framework to deal with such
computers and their use for remote access to organizational
vulnerabilities. The framework suggests specific courses of action
for two possible scenarios. When there is no present threat, a
networks.
proactive approach is suggested. When one or more threats are The proposed framework can help system administrators to
present, a reactive, matrix-based approach is suggested. Both assess the vulnerabilities associated with using mobile laptops
approaches offer a systematic methodology to address laptop to remotely access the local area networks (LAN) or wireless
vulnerabilities. A similar framework can be extended to address local area networks (WLAN). Once an assessment is made, the
security vulnerabilities of other mobile computing devices in network administrator can address such vulnerabilities in a
addition to notebooks and laptop computers. A real case scenario
systematic and efficient manner. Also, the framework suggests
from a network in a university college in the southeastern U.S. is
a step-by-step procedure to address such vulnerabilities when
used to illustrate the proposed framework.
the system is under attack, or when one or more threats are
Keywords - mobile computing; cybersecurity; vulnerability; present.
managerial approach The paper is organized as follows. First, a brief discussion
of vulnerabilities of mobile laptops and their use for remotely
I. INTRODUCTION accessing a given network is provided. The next section
Recent trends of globalization, outsourcing, off-shoring, discusses the modeling framework and presents the practical
and cloud computing have changed the structure of recommendations for system administrators. The framework
organizations and cyberspace. Information is no longer includes a proactive systematic approach to continuously
confined within the walls of an organization. Today’s evaluate the set of vulnerabilities and a reactive approach for
organizations are constantly allowing their suppliers to access dealing with vulnerabilities when one or more threats are
their supply chain management systems, customers to retrieve present. Finally, conclusions and several practical
product information from their electronic commerce systems, recommendations are provided
and their own employees to log on to the organizations’
intranet. Organizations use remote access to information II. VULNERABILITIES OF MOBILE COMPUTING
systems to streamline their business processes, become During the last two decades the popularity of notebooks and
operationally efficient, and to gain competitive advantage. laptops has increased significantly. They have been and will
However, the global reach of information systems has raised continue to be the computers of choice for individuals and
concerns over security and has made organizations more organizations. Forrester Research recently reported that laptop
vulnerable to security threats. sales in the U.S. overtook desktop sales 44 percent to 38
Organizations must pay special attention to cybersecurity percent in 2009 and 44 percent to 32 percent in 2010 [4]. The
vulnerabilities and ensure that their notebooks, laptops, and same report predicts that laptop sales will remain unchanged in
other mobile devices and networks are not compromised as a the 42-44 percent range for the next few years while desktop
result of this increase in mobility [1]. A recent study about sales will gradually decline to 18 percent in 2015. Laptops have
software vendors indicated that organizations lose become popular because they allow professionals and

1 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9 No. 3, March 2011
knowledge workers to access their networks when they are access, Internet access, and file transfer protocol (FTP) access.
travelling or from home offices and at the same time they offer Such actions create an environment for opening potential
storage and processing capabilities similar to, or even better harmful attachments, allowing potential unauthorized access to
than desktops. important files, potential for sniffing, session hijacking, IP
address spoofing, and denial of service attacks. In general,
The shift toward mobile computing is associated with a new using a laptop to access a WLAN is more susceptible to attacks
set of vulnerabilities for information systems. Mobile laptops because WLAN includes both the organization’s internal
are considered by most organizations as the greatest security network and the general public network segments. For
threat and the most difficult to maintain [3]. A survey example, WLANs can be susceptible to attacks such as traffic
published in 2006 indicated that in 27 percent of the cases, it analysis, eavesdropping, brute force attack, renegade access
took longer than 10 days to deploy critical patches to mobile points, and masquerading attacks.
laptops [3]. A timely and efficient response to laptop
vulnerabilities must be a major concern for organizations and System administrators and laptop users can address network
their system administrators. access vulnerabilities through several courses of action. They
can formulate and implement network access security policies,
Mobile computing vulnerabilities can be classified into require periodic change of login information and enforce a
three major categories: physical vulnerability, system policy for strong passwords, clearly define user privileges
vulnerability, and network access vulnerability. A brief (read, write, delete) and user access, and enforce secure setting
discussion of those categories is provided below along with a access and avoid access from open networks.
suggested course of actions.

A. Physical Vulnerability III. MANAGING VULNERABILITIES OF LAPTOP COMPUTERS
AND NETWORK ACCESS
Laptops are mobile computers and they travel with their
owners or users. There is a greater chance for laptops to be lost The identification of physical, system, and network access
or stolen in airports, hotels, and meeting auditoriums. Physical vulnerabilities allows the system administrator to prepare a
vulnerability is not only associated with the loss of hardware; it course of action to address these vulnerabilities. It is very
is also associated with the loss of valuable data and sensitive important that a continuously improvement plan is in place and
information. Another form of physical vulnerability occurs vulnerabilities are dealt with in a timely manner and preferably
when laptops are left open and unattended, which leads to before a threat occurs. Such an approach requires that security
exposure to sensitive information and documents and the perspective is shifted from technical to managerial. The main
ability for network access. goal of addressing vulnerabilities will be to improve business
resiliency and continuity [6].
System administrators must continuously raise awareness
about the importance of physical security and remind laptop A. Managing Vulnerabilities: No Present Threat
users of consequences of this vulnerability. In some cases, it is
necessary to secure the rooms or offices where the laptop is System administrators must continuously work to reduce
located and other times it is necessary to fasten the laptop to a the number of vulnerabilities present at any time during normal
non-movable object. business operations. Even when there is no immediate threat a
systematic, process based, proactive approach must be
followed. This approach has three major steps:
B. System Vulnerability
Laptop computer systems are as vulnerable as any other 1. Identify present vulnerabilities in the IT security area
computer system in the organization. A recent survey on laptop 2. Rate vulnerabilities based on the potential damage and
vulnerability assessment indicates that the most significant type likelihood of attack
of vulnerabilities are missing security patches and updates,
misapplied and outdated patches, outdated virus and spyware 3. Address vulnerabilities with specific course of action
definition files, configuration weaknesses that create exposures,
1) Identification of Vulnerabilities
and missing or deficient security applications, topologies and
processes [5]. Remote laptops can be physically accessed
easier than desktops. As such, non-secure laptop systems pose During normal business operations of the organizational
greater vulnerability than desktop systems. cyberspace, when there is no threat to the system, system
administrators must evaluate potential vulnerabilities of the
System administrators must prepare a schedule of updates system and among them, vulnerabilities of laptop computers
for security patches, antivirus programs, and other security and their access to the organizational network. The literature
programs. It is very important to follow the schedule and allow review and practical experience have identified a series of
users to update their systems as soon as a new update becomes vulnerabilities for any particular information system. Reference
available. [7] suggests a series of vulnerability categories related to
network access as shown in the first column of Table I.
C. Network Access Vulnerability System administrators must identify what vulnerabilities
The need to access LAN and WLAN using mobile laptops from the above list are present in his or her network. For those
creates the single most significant set of vulnerabilities for the vulnerabilities which are present the administrator must specify
organizational cyberspace. Laptops are used to provide e-mail

2 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9 No. 3, March 2011
any symptom(s), rating, and required action (s). This process is Timothy Parker is a systems administrator at the College
illustrated with a real case scenario as described below: of Business, an AACSB accredited institution in a regional
university in the southeastern U.S. The college has two
TABLE I. LIST OF VULNERABILITIES
computer laboratories, four computer classrooms, and many
lecturing podiums equipped with workstations and projectors.
Vulnerability Presen Symptoms Rating Action The college has an inventory of 78 laptops that are distributed
t? Required to faculty members for their research and teaching needs. The
Password cracking Yes Several faculty High Send a memo
members use with
college has several LANs, a secure WLAN, and an open
the same guidelines for wireless network. Faculty members use their laptops to access
password to strong student information, classroom information, and research files
access several passwords that are stored in several drives around the college’s LAN.
services such as and request Students also use their own laptops and mobile devices to
Blackboard, password access classroom information and other files located in the
Banner, and a changes.
shared server network.
with sensitive
Mr. Parker is aware that many faculty members use the
research
documents same password to access several services, including
Network and Blackboard, Banner, and servers with sensitive information.
system Students also use their laptops to access their records using an
information unsecured wireless network. Several laptops and desktops are
gathering infected due to students downloading harmful documents via
User enumeration
the Internet. Several new programs on the faculty laptops and
Backdoors,
Trojans and desktops need to be updated. Students use classroom and
remote controlling laboratory computers to access gaming Web sites. As Mr.
Gaining access to Yes Students are High Enforce Parker was walking through the building he noticed that some
remote using their secure wired faculty members had left their office open or unlocked with
connections and laptops to or wireless laptops already logged onto the network.
services access student connection to
records using sensitive data 2) Vulnerability Priority Ratings
the unsecured
wireless
network A system’s vulnerability rating represents a combination of
Privilege and user the potential damage a certain attack poses on the vulnerability
escalation and the attractiveness of the vulnerability in the eyes of an
Spoofing intruder. The following three vulnerability ratings are
Misconfigurations suggested:
Denial-of-service
(DoS) and buffer • High: This vulnerability is very attractive to the intruder and
overflows has high consequences if this vulnerability is exploited.
Viruses and Yes Several laptops High Update
worms and desktops antivirus
Mr. Parker has rated password cracking, gaining access to
are infected. programs and remote connections, presence of viruses and worms in this
scan and category.
clean the
infected • Moderate: This vulnerability is somewhat attractive to the
computers intruder and consequences if this vulnerability is exploited
Hardware specific are moderate. Mr. Parker has rated security policy
Software specific Yes Several new Low Update and violation in this category.
and updates programs need install new
to be updated in patches to • Low: This vulnerability is not very attractive to the intruder
the faculty improve and has low consequences if this vulnerability is exploited.
laptops and security
desktops.
Mr. Parker has rated software specific and updates in this
Security policy Yes Students use Modera Send a memo category.
violations classroom and te and remind
laboratory students and
3) Course of Actions
computers to faculty of
access gaming security Using the priority ratings identified in the previous step,
websites. Some policies Mr. Parker generates a working plan to address the
faculty related to this vulnerabilities in the College of Business. Specifically, Mr.
members leave vulnerability
open laptops in Parker must immediately send a memo with guidelines for
unlocked strong passwords and request password changes, enforce
offices secure wired or wireless connection to sensitive data, update
antivirus programs, scan, and clean the infected computers,

3 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9 No. 3, March 2011
send a memo and remind students and faculty of relevant 2) Evaluate the severity of each threat for each
security policies, and update and install new patches. vulnerability

B. Managing Vulnerabilities: Present Threat Each cell in Table II represents the severity (or risk) of a
When one or more threats are present, system given threat to a still existing vulnerability. High severity or
administrators must change the mode of operation from risk combinations are designated in red, moderate severity
proactive to reactive. When the system is under attack, a quick combinations are designated in yellow and low severity
evaluation of the threats and quick reaction to these threats is combinations in green. The interpretations of the severity
necessary. The reaction is immediate but still systematic, and ratings are provided below:
the following steps must be followed: Severity of this combination is high. The course of
1. Create a vulnerability-threat matrix action recommended to mitigate these
threats/vulnerabilities should be implemented
2. Evaluate the severity of each threat for each vulnerability immediately.
3. Address vulnerability-threat with specific course of action Severity of this combination is moderate. The course of
1) Create a vulnerability-threat matrix action recommended to mitigate these
threats/vulnerabilities should be implemented as soon
The vulnerability-threat assessments matrix can be utilized as possible.
with any information system or part of it. The matrix approach Severity of this combination is low. The course of
is often suggested in the literature [8] [9]. The matrix is used action recommended to mitigate these
to map the severity of a given threat with a given vulnerability threats/vulnerabilities will improve security, but is of
and to systematically generate an emergent and effective less urgency.
response. Table II is an illustration of this matrix from the
College of Business case. As shown in Table II, the spoofing attack is currently
presenting a moderate level of severity with regard to gaining
remote access to the network. In general, spoofing can be very
TABLE II. VULNERABILITY-THREAT MATRIX devastating for the organization (college) and the use of laptop
Unaddressed Threat 1: Threat 2: Action Required computers to access the network is a weakness for the system.
Vulnerabilities Spoofing New Virus is However, Mr. Parker is happy to see that his last memo on
Attack Spreading at security policy, the importance of strong passwords, and his
a High Rate action to request password changes have transformed this
Gaining access Enforce secure
to remote wired or wireless
potentially high risk threat-vulnerability combination into a
connections and connection to moderate level. On the other hand, the spread of new viruses is
services sensitive data causing significant damage to the laptops and other machines
Viruses and Update antivirus that are already infected or which do not have up-to-date
worms programs and scan antivirus protections.
and clean the
infected computers 3) Address vulnerability-threat with specific course of
Software Update and install action
specific and new patches to
updates improve security
Based on the findings from the previous step, system
administrators need to identify the immediate course of action
Mr. Parker has addressed several vulnerabilities but is still to address the most severe vulnerability-threat. Specifically,
working on enforcing secure connection, performing the latest Mr. Parker must update antivirus programs and scan and clean
update to the antivirus programs, and scanning and cleaning the all the infected laptop and desktop computers. Simultaneously,
several infected computers. Suddenly, Mr. Parker is made he needs to install new patches to improve security for the rest
aware of two security threats. First, a spoofing e-mail is of the network. Additionally, Mr. Parker must address the
circulating among the faculty members’ and students’ moderate vulnerability-threat combination by enhancing the
electronic mailboxes. The e-mail asks recipients to login to a security of the wired and wireless networks.
Web site and verify their login information or their e-mail
service will be interrupted. Second, several faculty members IV. SUMMARY AND RECOMMENDATIONS
are reporting that many computers in the computer lab have
stopped responding due to what seems to be a Trojan attack. As Notebooks and laptops have become the computers of
the first step, Mr. Parker builds the vulnerability-threat matrix choice for professionals and managers who want to access their
as shown in Table 2. Only the unaddressed vulnerabilities are organizational networks while traveling or while working from
listed in this table along with their typical course of actions. home. With this popularity they also offer the greatest security
challenges for system administrators. Laptops and their use to
access organizational networks produce three major
vulnerability categories: physical, system, and network access.

4 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9 No. 3, March 2011
The paper discuses these vulnerabilities and offers a framework http://www.edtechmag.com/higher/docs/2008/09/mobile-computing-
for addressing them. security.pdf.
[2] R. Telang, R. and S. Wattal, “An empriical analysis of the impact of
In general, there are two scenarios under which a system software vulnerability announcement on form stock price, “ in IEEE
administrator can address the vulnerabilities. The first scenario Transactions on Software Engineering, Vol 33 (8), pp. 544-557, 2007.
assumes no presence of a given threat and is designed to [3] B. Bosen, “Vulnerability management survey” in Trusted Strategies,
provide a systematic and proactive course of action to 2006, Retrieved February 7, from 2011. http://www.trusted
strategies.com/ papers/vulnerability_management_survey.pdf.
continuously improve the security of the laptops and their use
[4] E. Schonfeld, “Forrester projects Tablets will outsell Netbooks by 2012,
to access organizational LANs or WLANs. The scenario Desktops by 2013” June 2010, Retrived February 9, 2011 from
suggests a course of action based on a vulnerability rating http://techcrunch.com/2010/06/17/forrester-tablets-outsell-netbooks/
system. The vulnerabilities are rated based on two factors: the [5] Fiberlink, “Laptop vulnerability sssesment service,” 2011, retrieved n
degree of attractiveness to a potential intruder and the February 8, 2011 from http://feeneywireless.com/fetchdoc.php?docID
consequences/impact of the vulnerability for the organization. =90856300.
[6] J. Allen, J. “The art of information security governance” in Qatar
The second scenario assumes the presence of one or more information security forum, 2008, Software Engineering Institute,
security threats. This scenario is designed to offer a reactive, retrieved on February 8, 2011 from http://www.cert.org/archive/pdf/
but systematic course of action. A matrix is designed, and in QISF_Allen_022408.pdf.
each cell of the matrix, the severity of a vulnerability-threat [7] H. S. Venter, and J. H. Eloff, “Vulnerabilities categories for intrusion
combination is represented with a color coded sign. Again, a detection systems in Computers & Security, Vol. 21 (7), pp. 617-619,
2002.
course of action is suggested starting with the most severe
combinations, followed by moderate combinations, and ending [8] S. Goel and V. Chen, “Information security risk analysis–a matrix-based
approach, 2005, retrieved on February 7, 2011 from
with the low risk combinations. http://www.albany.edu/~goel/publications/goelchen2005.pdf.
[9] N. A. Renfroe and J. L. Smith, “Threat/vulnerability assessments and
V. CONCLUSIONS risk analysis” November 2010, retrived on February 7, 2011
fromhttp://www.wbdg.org/resources/riskanalysis.php.
This paper offers a managerial framework for addressing
laptop physical, system, and network access vulnerabilities. AUTHORS PROFILE
The purpose of the framework is to assist system administrators
Arben Asllani is a Post Doctoral Fellow in Cybersecurity at the Center
to create effective action plans to deal with such vulnerabilities. for Secusrity Studies at the University of Maryland University College
A proactive approach to eliminating vulnerabilities is suggested (UMUC) and a UC Foundation Professor of Management at the
and a step-by-step methodology is offered. When security University of Tennessee at Chattanooga. He has published over 24
threats are present, a matrix-based approach is suggested. The journal articles and presented and published over twenty conference
proceedings. His most recent research has been published in such
matrix can help the system administrator identify the most journals as Omega, European Journal of Operational Research,
severe attack/vulnerability combination and mitigate the risk of Knowledge Management, and Computers & Industrial Engineering.
such threats. The matrix based approach is a reactive approach
but it is necessary to guide the system administrator when the Amjad Ali is the Director of the Center for Security Studies and a
networks or laptop computers are under attack. A real case Professor of Cybersecurity at University of Maryland University
scenario from a university college is used to illustrate the College. He played a significant role in the design and launch of
framework. The suggested framework is not limited to the use UMUC’s cybersecurity programs. He teaches graduate level courses in
of laptop computers; it can be used by organizations to monitor the area of cybersecurity and technology management. He has served as
a panelist and a presenter in major conferences and seminars on the
vulnerabilities in other areas of organizational cyberspace. topics of cybersecurity and innovation management. He is a member of
the Maryland Higher Education Commission (MHEC) Cybesecurity
REFERENCES Advisory Council, providing advice and help on how MHEC can
respond best to the higher education needs of the growing cybersecurity
[1] CDW-G (White Paper), “Mobile computing security: protecting data on workforce.
devices roaming on the perimeter,” Retrieved March 7, 2011, from:

5 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Reduction of PAPR for OFDM Downlink
and IFDMA Uplink Wireless Transmissions
Bader Hamad Alhasson Mohammad A. Matin
Department of Electrical and Computer Engineering Department of Electrical and Computer Engineering
University of Denver University of Denver
Denver, USA Denver, USA

Abstract-- One of the major drawbacks of OFDM is the high ad-hoc mode or access point for current wide use. In 1997
peak-to-average power ratio (PAPR) of the transmitted signals. WLAN standard – IEEE 802.11, also known as Wi-Fi, was
In this paper, we propose a novel low complexity clipping scheme first developed with speeds of up to 2 Mbps [2]. At present,
applicable to Interleaved-FDMA uplink and OFDM downlink
systems for PAPR reduction. We show the performance of PAPR
WLANs are capable of offering speeds up-to 600 Mbps for the
of the proposed Interleaved-FDMA scheme is better than IEEE 802.11n utilizing OFDM as a modulation technique in
traditional OFDMA for uplink transmission system. Our the 2.4 GHz and 5 GHz license-free industrial, scientific and
reduction of PAPR is 53% when IFDMA is used instead of medical (ISM) bands. It is important to note that WLANs do
OFDMA in the uplink transmission. We also examine an not offer the type of mobility, which mobile systems offer. In
important trade-off relationship between clipping distortion and our previous work, we modeled a mix of low mobility
quantization noise when the clipping scheme is used for OFDM 1.8mph, and high mobility, 75mph with a delay spread that is
downlink systems. Our results show that we were able to reduce constantly slighter than the guard time of the OFDM symbol
the PAP ratio by 50% and reduce the out-of-band radiation to predict complex channel gains by the user by means of
caused by clipping for OFDM downlink transmission system.
reserved pilot subcarriers [3].
Orthogonal frequency division multiplexing (OFDM) is a
broadband multicarrier modulation scheme. Research on
Keywords-component-- Signal to quantization noise ratio multi-carrier transmission started to be an interesting research
(SQNR);Localized-frequency-division-multiple-access (LFDMA); area [4-6]. OFDM modulation scheme leads to better
interleaved-frequency-division-multiple-access (IFDMA); peak-
performance than a single carrier scheme over wireless
to-average power ratio (PAPR); clipping ratio (CR); single
carrier frequency division multiple access (SC-FDMA). channels since OFDM uses a large number of orthogonal,
narrowband sub-carrier that are transmitted simultaneously in
parallel. We investigated the channel capacity and bit error
I. INTRODUCTION rate of MIMO-OFDM [7]. The use of OFDM scheme is the
solution to the increase demand for future bandwidth-hungry
Wireless communication has experienced an incredible growth
wireless applications [8]. Some of the wireless technologies
in the last decade. Two decades ago the number of mobile
using OFDM are Long-Term Evolution (LTE). LTE is the
subscribers was less than 1% of the world’s population [1]. In
standard for 4G cellular technology, ARIB MMAC in Japan
2001, the number of mobile subscribers was 16% of the
have adopted the OFDM transmission technology as a
world’s population [1]. By the end of 2001 the number of
physical layer for future broadband WLAN systems, ETSI
countries worldwide having a mobile network has
BRAN in Europe and Wireless local-area networks (LANs)
tremendously increased from just 3% to over 90% [2]. In
such as Wi-Fi. Due to the robustness of OFDM systems
reality the number of mobile subscribers worldwide exceeded
against multipath fading, the integration of OFDM technology
the number of fixed-line subscribers in 2002 [2]. As of 2010
and radio over fiber (RoF) technology made it possible to
the number of mobile subscribers was around 73% of the
transform the high speed RF signal to the optical signal
world’s population which is around to 5 billion mobile
utilizing the optical fibers with broad bandwidth [9].
subscribers [1].
Nevertheless, OFDM suffers from high peak to average power
ratio (PAPR) in both the uplink and downlink which results in
In addition to mobile phones WLAN has experienced a rapid
making the OFDM signal a complex signal [10].
growth during the last decade. IEEE 802.11 a/b/g/n is a set of
standards that specify the physical and data link layers in

6 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

N C 1 N C 1
x (t )  
K 0
x k (t )  a
K 0
m
e j 2  k  ft (1)

Max linear limit
Where xk (t) is the kth modulated subcarrier at a
frequency f k  k .f . The modulation symbol a k ) is
(m

Figure.1 Fresnel diagram illustrating the PAPR issue
applied to the kth subcarrier during the mth OFDM
Figure.1 shows a constructive addition of subcarriers on a interval which is mTu  t  ( m  1)TU . Therefore, during
random basis which causes the peak-to-average power ratio
problem. The outcome of high PAPR on the transmitted each OFDM symbol interval transmission, N C modulation
OFDM symbols results in two disadvantages high bit error symbols are transmitted in parallel. The modulation symbols
rate and inference between adjacent channels. This would are dependent on the use of this technology and can be any
imply the need for linear amplification. The consequence of form of modulation such as 16QAM, 64QAM or QPSK. The
linear amplification is more power consumption. This has choice of which modulation scheme to implement varies
been an obstacle that limits the optimal use of OFDM as a
modulation and demodulation technique [11-14]. The problem depending on the environment and application.
of PARP affects the uplink and downlink channels differently.
On the downlink, it’s simple to overcome this problem by the
m
a0 e j 2f 0t x (t )
use of power amplifiers and distinguished PAPR reduction 0

methods. These reduction methods can’t be applied to the m m m
a0 , a1 ,..., a NC 1
a1m e j 2f1t x (t )
uplink due to their difficulty in low processing power devices Serial to 1

such as mobile devices. On the uplink, it is important to parallel x (t )
+
reduce the cost of power amplifiers as well.
j 2f N C 1t
PAPR reduction schemes have been studied for years [15-18]. m e
a N C 1 x N C 1 (t )
Some of the PAPR reduction techniques are: Coding
techniques which can reduce PAPR at the expense of f k  kf
bandwidth efficiency and increase in complexity [19-20]. The
probabilistic technique which includes SLM, PTS, TR and TI
can also reduce PAPR; however; suffers from complexity and Figure 3 OFDM modulation valid for time interval
spectral efficiency for large number of subcarriers [21-22]. mT u  t  ( m  1 ) T U .

We perform an analysis on a low complexity clipping and Subcarriers spacing range hundreds of kHz to a small number
filtering scheme to reduce both the PAPR and the out-of-band- of kHz depending on the environment of operation. Once the
radiation caused by the clipping distortion in downlink spacing between subcarriers has been specified, then the
systems. It was also shown that a SC-FDMA system with choice of how many subcarriers to be transmitted in parallel
Interleaved-FDMA or Localized-FDMA performs better than has to be done. It is important to note that allocation of the
Orthogonal-FDMA in the uplink transmission. number of subcarriers is dependent on the transmission
bandwidth. For instance, LTE uses 15 kHz as the basic
II. SYSTEM MODEL spacing with a 600 subcarriers assuming the operation is in the
10 MHZ spectrum.
Let us consider two modulated OFDM subcarriers x k 1 (t ) and
x k 2 (t ) . The two signals are orthogonal over the time period
IFFT Clipping Filtering
mTu  t  (m  1)TU
( m 1)Tu


Figure 2 Clipping and Filtering at the Transmitter of OFDM system *
x k 1 ( t ) x k 2 ( t ) dt
In complex baseband, an OFDM signal x(t ) during time mT u

interval mTu  t  ( m  1)TU can be expressed as

7 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

( m 1) Tu add up coherently with identical phases. The largest PAPR

a
j 2 k1 f  j 2 k 2  f
*
a e
k1 k 2 e dt  0 happens randomly with a very low probability. The main
interest is actually in the probability of the occurrence of high
mTu signal power. This high signal power is out of the linear range
of high power amplifiers. The probability PAPR is below a
for k1  k 2 (2) certain threshold can be expressed as:

Therefore, OFDM transmission can be expressed as the
P(PAPR z)  F ( z) N  (1  exp(z))N (6)
modulation of a set of orthogonal functions  k (t ) , where
j 2  k  ft Equation (6) holds for samples that are mutually uncorrelated;
 k (t ) = e 0  t  Tu , however; when over sampling is applied then it doesn’t hold.
This is due to the fact that a sampled signal doesn’t certainly
0 otherwise (3)
include the maximum point of the original continuous time
signal. Nevertheless, it is important to note that it is difficult to
Pilot derive the exact cumulative distribution function for the peak
power distribution. The following simplified proposed PAP
User A User B distribution will be used:

F ( z ) N  (1  exp( z 2 ))N (7)

Where  has been found by fitting the theoretical CDF into
the actual one. From our simulation, it was shown that  =2.8
Frequency is suitable for adequately a large number of subcarriers.
Guard Band

The theoretical and simulated curves are plotted in Figure 5
for different number of subcarriers. As N decreases, the
Figure 4. OFDM available bandwidth is divided into subcarriers that are deviation between the obtained simulation and theoretical
mathematically orthogonal to each other
results increases, which indicates that equation (7) is quite
accurate for N>256. It is worth noting that equation (6) is
III. DISTRIBUTION of THE PAP RARIO more accurate for large CDF values as shown in Figure 5.

0
10
The complex baseband signal for one OFDM symbol can be Theoretical
rewritten as: Simulated

N=16
1 N
x(t )   an exp( j n t ) N=1024

N n1 (4) N=32
CCDF

-1
10
N=64
Where N is the number of subcarriers and an are the
modulating symbols. From the central limit theorem, we can N=128
assume that the real and imaginary parts of the time domain
complex OFDM signal x(t ) have a Gaussian distribution for N=512

a large number of subcarriers. Therefore, the amplitude of the
OFDM signal x(t ) follows a Rayleigh distribution, whereas -2
10
2 3 4 5 6 7 8 9 10 11
power follows a central chi-square distribution with the PAPR[dB]
cumulative distribution expressed as:
Figure 5 OFDM system with N-point FFT. CCDFs of signal PAP ratio with
N=16, 32, 64, 128 and 1024. Solid lines are calculated; dotted lines are
F ( z)  1  e z (5) simulated.

OFDM system with a certain number of subcarriers suffers
from maximum power which arises when all of the subcarriers

8 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

IV. CLIPPING AND SIGNAL TO QUANTIZATION
NOISE RATIO The performance of any PAPR reduction scheme is evaluated
based on out-of-band radiation, in-band ripple, distribution of
An OFDM signal has the tendency to have a large peak to PAPR and the BER performance [23].
average power ratio when each subcarrier by chance has the
highest amplitude and identical phases at the same time. The V. SIMULATION AND RESULTS
likelihood of such event is rare yet it does occur. As the
number of subcarriers increase, the maximum power increases To evaluate the performance of the clipping and filtering
as shown in Figure 5. The probability of that maximum power method used in our simulation, the following parameters were
signal actually decreases as N increases. This is due to the used to in the simulation.
statistical magnitude distribution of the time-domain OFDM
signal. Table I. Simulation parameters
The simplest approach to reduce the PAP ratio is to clip the
amplitude of the signal to a desired maximum level. Although N 256
clipping is the simplest method, in our method it enhances the Clipping Ratio 1.4
signal to quantization noise ratio (SQNR) in the conversion
from analog to digital. Carrier frequency 5 MHz
Modulation QPSK
As the clipping threshold increases, clipping distortion Sampling frequency 10 MHz
decreases at the expense of PAPR and quantization noise. On
the other hand as the clipping threshold decreases, PAPR and Bandwidth 1MHz
quantization noise decrease at the expense of clipping Guard interval samples 32
distortion. Therefore, it is important to take into consideration
this trade-off relationship between clipping distortion and
quantization noise when picking the number of bits for 0.4
quantization and the clipping threshold.
0.3
abs(x ”[m ])

Figure 6 shows the SQNR values of OFDM signal quantized
0.2
with 5, 6, 7, 8 bits against the clipping threshold and N=128.
The optimal clipping threshold to maximize the signal to 0.1
quantization noise ratio fluctuates with the quantization level;
however; we can see that the maximum points are 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
approximately around 3.5 of the normalized clipping time
threshold. Clipping distortion is more significant to the left of
0
the maximum points due to the low threshold of clipping
whereas the clipping distortion is not as significant to the right
of the maximum points where the clipping threshold is higher.
P S D[dB ]

-50
45

40 8bits
-100
-5 -4 -3 -2 -1 0 1 2 3 4
Hz 6
x 10
35 7bits
Figure 7 Baseband signal
SQNR[dB]

30 Figure 7 shows the power spectral density of oversampled
6bits
baseband signal. This is the output of IFFT. Let x(s) be the
output of IFFT. Then the output of IFFT can be expressed
25
mathematically as:
5bits
L . N 1
1
20 x(s) 
L.N
 X ( k ). e
k 0
2  js  fk / L . N
, s  0 ,1,... NL  1
With
15
2 2.5 3 3.5 4 4.5
clipping level)
5 5.5 6 6.5 7
X ( k ) = X ( k ) , for 0  k< N/2 and NL-N/2< k <NL
0, otherwise (8)
Figure 6 Clipping threshold against SQNR of quantized
OFDM signal. N=128

9 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Where L, f , N and X ( k ) represent the oversampling factor, Clipping Ratio=1.4
0.04
the subcarrier spacing, the number of subcarriers and the
0.03
symbol carried by subcarrier k, respectively.

pdf
Gaussian distribution 0.02
0.08
0.01
0.06
0
pdf

0.04 -0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1
x
0.02 Out-of-band radiation
0
reduction after filtering
0
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
x

PSD[dB]
-50
0
PSD[dB]

-100
-50 -5 -4 -3 -2 -1 0 1 2 3 4
Hz 6
x 10
Figure 10 Clipped and filtered passband signal
-100
-5 -4 -3 -2 -1 0 1 2 3 4
Hz 6
x 10
Figure 8 Baseband signal The out-of-band radiation can be seen from Figure 9 and 10. It
is clear that the out-of-band radiation increases after clipping;
Figure 8 shows the power spectral density and a histogram of however; it decreases after filtering and shows a peak value
the baseband signal without clipping and filtering. We can see beyond the clipping threshold implying a slight peak re-
the power density function shows a Gaussian distribution of growth in PAPR after filtering as shown in Figure 10. To
the signal. complete the evaluation of clipping and filtering then we have
to look at the BER performance when the clipping ratio varies.
Clipping Ratio=1.4
0.2

0.15 0
10
pdf

0.1

0.05
Clipped
Clipped & filtered
CCDF

0 -1
-0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 10 Unclipped
x
Out-of-band radiation
0
due to clipping
PSD[dB]

-2
-50 10
2 4 6 8 10 12 14 16
PAPR[dB]
-100
-5 -4 -3 -2 -1 0 1 2 3 4 0
Hz 6 10
x 10
Figure 9 Clipped passband signal

Clipping and filtering OFDM has been studied [23]; however;
-2
these techniques reduce PARP at the expense of increased
BER

10
system complexity and a high peak re-growth. Figure 9 shows
the level of Out-of-band radiation increases as the OFDM Unclipped
signal passes through a nonlinear device. An OFDM
transmitter emits out-of-band radiation when a set of -4
subcarriers are modulated. Our results show less out-of-band 10
0 1 2 3 4 5 6 7 8 9 10
power emission compared to traditional OFDM by the use of
the low complexity clipping and filtering technique. EbNo

Figure 11 (a) PAPR distribution (b) BER performance

10 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

It can be seen from Figure 11(a) as the clipping ratio increases
0
from right to left, the PAPR decreases dramatically after 10
clipping and increases slightly after filtering. The simulation
result in Figure 11 (b) shows that the performance of BER is
better as the clipping ratio increases. Unlike OFDM used for -1
downlink transmission, SC-FDMA is utilized in the uplink 10

transmission where subcarriers are separated and designated

Pr(PAPR>PAPR 0 )
for several mobile units. Each unit utilizes a number of
subcarriers, let N denote the number of subcarriers assigned 10
-2
unit
to each unit for uplink transmission. The effectiveness of
reduction in PAPR is greatly influenced by the technique in
the method utilized to assign N to each unit [24]. 10
-3
unit
Orthogonal-FDMA
Discrete Fourier Transform (DFT) spreading technique is a
Localized-FDMA
promising solution to reduce PAPR because of it’s superiority
Inter leaved-FDMA
in PAPR reduction performance compared to block coding, -4
10
Selective Mapping (SLM), Partial Transmit Sequence (PTS) 0 2 4 6 8 10 12
and Tone Reservation (TR) [25-26]. SC-FDMA and OFDMA PAPR in dB
are both multiple-access versions of OFDM. There are two Figure 12 (b) 16 QAM
subcarrier mapping schemes in single carrier frequency
division multiple access (SC-FDMA) to allocate subcarriers The three figures of 12 show that when the single carrier is
between units: Distributed FDMA and Localized FDMA. mapped either by LDMA or DFDMA, it outperforms OFDMA
due to the fact that in an uplink transmission, mobile terminals
work differently then a base station in terms of power
0
10 amplification. In the uplink transmission PAPR is more of a
significant problem then on the downlink due to the type and
capability of the amplifiers used in base station and mobile
-1
devices. For instance, when a mobile circuit’s amplifier
10 operates in the non-linear region due to PAPR, the mobile
devise would consume more power and become less power
P r(P A P R> P A P R 0 )

efficient whereas base stations don’t suffer from this
-2
consequence. Therefore, OFDM works better in the downlink
10 transmission in terms of PAPR.
0
10

-3
10
Orthogonal-FDMA
Localized-FDMA -1
10
Interleaved-FDMA
-4
P r(P A P R> P A P R 0 )

10
0 2 4 6 8 10 12
PAPR in dB
-2
10
Figure 12 (a) QPSK

Figure 12 show the performance of PAPR while the number of
subcarriers is 128 and the number of subcarriers assigned to -3
10
each unit or mobile device is 32. This simulation helps in Orthogonal-FDMA
evaluating the performance of PAPR with different mapping Localized-FDMA
schemes and modulation techniques. In LFDMA each user Interleaved-FDMA
transmission is localized in the frequency domain where in the -4
10
DFDMA each user transmission is spread over the entire 0 2 4 6 8 10 12
PAPR in dB
frequency band making it less sensitive to frequency errors
and diversifies frequency. Figure 12 (c) 64 QAM
Our results show the effect of using Discrete Fourier
Transform spreading technique to reduce PAPR for OFDMA,
LDMA and OFDMA with N=128 and N =32. A comparison
unit

11 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

is shown in Figure 12 a,b and c utilizing different modulation REFERENCES
schemes. The reduction in PAPR is significant when DFT is
used. For instance, Figure 12(b) where Orthogonal-FDMA, [1] ITU, “World Telecommunication Development Report 2002:
Localized-FDMA and Interleaved-FDMA have the values of Reinventing Telecoms”, March 2002.
5.9 dB, 9 dB and 11 dB, respectively. The reduction of PAPR [2] Anthony Ng’oma, “Radio-over-Fibre Technology for Broadband
Wireless Communication Systems”, June 2005.
in IFDMA utilizing the DFT-spreading technique compared to [3] Bader Alhasson, and M. Matin “The challenge of scheduling user
OFDMA without the use of DFT is about 53 percent. Such transmissions on the downlink of a long-term evolution (LTE)
reduction is significant in the performance of PAPR. Based on cellular communication system”, Proc. SPIE, Vol. 7797, 779719,
the simulation results in Figure 12 we can see that single Sep 2010.
[4] H. Atarashi, S. Abeta, and M. Sawahashi, “Variable spreading
carrier frequency division multiple access systems with factor orthogonal frequency and code division multiplexing
Interleaved-FDMA and Localized-FDMA perform better than (VSF-OFCDM) for broadband packet wireless access,” IEICE
OFDMA in the uplink transmission. Although Interleaved- Trans. Commun., vol. E86-B, pp. 291-299, Jan. 2003.
FDMA performs better than OFDMA and LFDMA, LFDMA [5] R. Kimura and F. Adachi, “Comparison of OFDM and multicode
MC-CDMA in a frequency selective fading channel, ” IEE
is preferred due to the fact that assigning subcarriers over the Electronics Letters, vol. 39, no.3, pp. 317-318, Feb. 2003.
whole band of IFDMA is complicated while LFDMA doesn’t [6] Z Wang and G. B. Giannakis, “Complex-field Coding for OFDM
require the insertion of pilots of guard bands. over Fading Wireless Channels,” IEEE Trans. Inform. Theory, vol.
49, pp.707-720, Mar. 2003.
[7] Alhasson Bader, Bloul A., Li X., and Matin M. A: “LTE-advanced
VI. CONCLUSION
MIMO uplink for mobile system” Proc. SPIE, Vol. 7797, 77971A,
2010.
We have shown the importance of the trade-off relationship [8] L. Mehedy, M. Bakaul, A. Nirmalathas, "115.2 Gb/s optical OFDM
between clipping distortion and quantization noise. Our results transmission with 4 bit/s/Hz spectral efficiency using IEEE
802.11a OFDM PHY," in proc. the 14th OptoElectronics and
show that as the clipping threshold increases, clipping
Communications Conference, 2009 (OECC 2009), July 2009.
distortion decreases at the expense of PAPR and quantization [9] Alhasson Bader, Bloul A., and Matin M. A.: “Dispersion and
noise. On the other hand as the clipping threshold decreases, Nonlinear Effects in OFDM-RoF system”, SPIE, Vol. 7797,
PAPR and quantization noise decrease at the cost of clipping 779704, 2010.
[10] J. Tellado, “Multicarrier transmission with low PAR,” Ph.D.
distortion. Therefore, it is important to take into consideration
dissertation,Stanford Univ., Stanford, CA, 1998.
this trade-off relationship between clipping distortion and [11] Z.-Q. Luo and W. Yu, “An introduction to convex optimization
quantization noise when picking the number of bits for for communications and signal processing," IEEE J. Sel. Areas
quantization and the clipping threshold. We showed that Communication, vol. 24, no. 8, pp. 1426-1438, Aug. 2006.
[12] J. Tellado, “Peak to average power reduction for multicarrier
clipping decreases the amplitude to a desired maximum power
modulation," Ph.D. dissertation, Stanford University, Stanford,
level and the outcome signal suffers from out-of-band USA, 2000.
radiation as a result of clipping distortion which is eliminated [13] A. Aggarwal and T. Meng, “Minimizing the peak-to-average power
by filtering at the expense of a slight peak re-growth in ratio of OFDM signals using convex optimization," IEEE Trans.
Signal Process., vol. 54, no. 8, pp. 3099-3110, Aug. 2006.
amplitude due to filtering. It was also shown that a SC-FDMA [14] Y.-C. Wang and K.-C. Yi, “Convex optimization method for
system with Interleaved-FDMA or Localized-FDMA performs quasiconstant peak-to-average power ratio of OFDM signals,"
better than Orthogonal-FDMA in the uplink transmission IEEE Signal Process. Lett., vol. 16, no. 6, pp. 509-512, June 2009.
where transmitter power efficiency is of great importance in [15] S. H. Wang and C. P. Li, “A low-complexity PAPR reduction
scheme for SFBC MIMO-OFDM systems,” IEEE Signal Process.
the uplink. Lett., vol. 16, no. 11, pp. 941–944, Nov. 2009.
LFDMA and IFDMA result in lower average power values, [16] J. Hou, J. Ge, D. Zhai, and J. Li, “Peak-to-average power ratio
due to the fact that OFDM and OFDMA map their input bits reduction of OFDM signals with nonlinear companding scheme,”
straight to frequency symbols where LFDMA and IFDMA IEEE Trans. Broadcast., vol. 56, no. 2, pp. 258–262, Jun. 2010.
[17] T. Jaing, W. Xiang, P. C. Richardson, D. Qu, and G. Zhu, “On the
map their input bits to time symbols. We conclude that single nonlinear companding transform for reduction in PAPR of MCM,”
carrier-FDMA is a better choice on the uplink transmission for IEEE Trans. Wireless Commun., vol. 6, no. 6, pp. 2017–2021, Jun.
cellular systems. Our conclusion is based on the better 2007.
efficiency due to low PAPR and on the lower sensitivity to [18] S. H. Han and J. H. Lee, “An overview of peak-to-average power
ratio reduction techniques for multicarrier transmission,” IEEE
frequency offset since SC-FDMA has a maximum of two Wireless Commun., vol. 12, pp. 56–65, Apr. 2005.
adjacent users. [19] Wilkinson, T.A. and Jones, A.E . Minimization of the peak-to-
mean envelope power ratio of multicarrier transmission scheme by
VII. FUTURE WORK block coding. IEEE VTC’95, Chicago, vol. 2,pp. 825-829. July,
1995.
[20] Park, M.H. PAPR reduction in OFDM transmission using
We would like to further investigate the effect of PAPR in Hadamard transform. IEEE ICC’00, vol.1, pp. 430-433. 2000
MIMO-OFDM systems. [21] Bauml, R.W., Fischer, R.F.H., and Huber, J.B. Reducing the peak
to average power ratioof multicarrier modulation by selective
mapping. Electron. Lett., 32(22), 2056-2057. 1996.
[22] Muller, S.H and Huber, J.B. a novel peak power reduction scheme
for OFDM. PIMRC, vol. 3,pp. 1090-1094. 1997.
[23] Li, X. and Cimini, L.J. Effects of clipping and filtering on the
performance of OFDM. IEEE Commun. Letter, 2(20), pp. 131-133
1998.

12 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

[24] Myung, H.G., Lim, J., and Goodman, “Single Carrier FDMA
forUplink Wireless Transmission,” IEEE Veh. Technol. Mag.,
1(3), 30-38. 2006.
[25] H. G. Myung, J. Lim, and D. J. Goodman, “Single Carrier FDMA
for Uplink Wireless Transmission,” IEEE Vehicular Technology
Mag., vol.1, no. 3, pp. 30 – 38, Sep. 2006.
[26] H. G. Myung and David J. Goodman, Single Carrier FDMA”,
WILEY, 2008.
[27] Cho, kim, Yang & Kang. MIMO-OFDMWireless Communications
with MATLAB. IEEE Press 2010.
[28] Bloul A., Mohseni S., Alhasson Bader, Ayad M., and Matin M. A.:
“Simulation of OFDM technique for wireless communication
systems”, Proc. SPIE, Vol. 7797, 77971B, 2010.

AUTHORS PROFILE

Bader Hamad Alhasson is a PhD candidate
from the University of Denver. He received a
bachelor degree in Electrical Engineering
(EE) in 2003 from the University of
Colorado at Denver (UCD) in the United
States, a Master’s of Science in EE and a
Master’s of Business Administration (MBA)
in 2007 from UCD. His primary research
interest is in the transmission and reception
of radio over fiber (RoF) utilizing OFDM.

Dr. Mohammad Abdul Matin, Associate
Professor of Electrical and Computer
Engineering, in the School of Engineering and
Computer Science, University of Denver. He
is a Senior Member of IEEE and member of
SPIE, OSA, ASEE and Sigma Xi. His
research interest is in Optoelectronic Devices
(such as Sensors and Photovoltaic)
RoF, URoF, Digital, Optical & Bio-Medical
Signal & image Processing
Engineering Management and Pedagogy in
Engineering Education.

13 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Evolution Prediction of the Aortic Diameter Based on
the Thrombus Signal from MR Images on Small
Abdominal Aortic Aneurysms
A. Suhendra1, C.M. Karyati2, A.Muslim3, A.B. Mutiara4
Faculty of Computer Science and Information Technology, Gunadarma University
Jl. Margonda Raya No.100, Depok 16424, Indonesia
1,2,3,4
{adang,csyarah,amuslim,amutiara}@staff.gunadarma.ac.id

Abstract—The paper is about studying the T1 and T2 from parts of the human body. The human blood pressure will refer
Magnetic Resonance (MR) Images examination for the existence to how much pressure in the arteries that brings blood to all
of thrombus in patient with Small Abdominal Aortic Aneurysms cells of the human body through the delicate vessels
(SAAA) in order to know whether thrombus signal has (capillaries) which then will return to the heart through blood
correlation with the evolution of aortic diameter enlargement, vessels and takes oxygen through the lungs. There are a little
which then can be used to predict the risk of rupture of aortic description of the aorta which will be discussed further in this
wall. Data were derived from 16 patients with SAAA, whereas study. It could be imagined if there are any damage to the
MR images obtained from 3T imager (Trio TIM, Siemens human aorta would result in abnormalities in blood flow in the
Medical Solution, Germany), which came from: the study of
human body. In the following image, we can see the anatomy
anatomy, cine-MR images, pictures T1/T2, blood flow images,
and images after injection of contrast agents. The surface area of
of the aorta and the arteries (figure 1) :
the aorta and luminal are determined by tracing manually, which
can be used to determine the surface area of thrombus. The
maximum diameter of the aorta are automatically obtained from
manual tracing on T1 images. The parameters to study the
thrombus signal are the mean, median, standard deviation,
skewness and kurtosis. Each parameter is calculated on the area
of thrombus, while for normalization we implement the signal in
the muscles. All parameters are compared to evolution of aortic
diameter. We found 13 out of 16 patients with SAAA have
thrombus. But there is no correlation between thrombus signals
and maximum diameter (mean (r = 0.318), median (r = 0.318),
skewness (r = 0.304)), or even with maksimum evolution diameter
(mean (r=0.512)). As the conclusion is the comparation between
mathematical and visual calculation of thrombus categories
reached 81% similar, but thrombus signal itself cannot be used to
Figure 1. Anatomy of the aorta [1]
predict the evolution of aortic diameter.
The Studies of human aorta have been conducted and
Keywords-component; Thrombus signal; evolution of aortic
diameter; T1 and T2 weighted images; Small Abdominal Aortic successfully detected abnormalities in the aortic wall, both at
Aneurysms. the thoracic or abdominal aortas [1,2]. In general, the swelling
of the aortic wall is very elastic, therefore if the swelling is
occur then aortic wall will not be able to shrink back and it will
I. INTRODUCTION be broken without being able to predict when the rupture risk
Aorta is the larger artery that delivers blood from the heart of the aortic wall. It could be in the risk of patient death.
of human beings throughout the body. In this way, the human
An Abdominal Aortic Aneurysm, also called AAA, is a
blood flow will go through some branch, for example, that led
bulging area in the wall of the aorta which is causing of an
to the arm (subclavian arteries), heading toward of the head
abnormal widening or ballooning until greater than 50 percent
(carotid arteries), and headed toward of the chest (thoracic
of the normal diameter. The the swelling of the aortic wall
aorta), then toward of the diaphragm to the stomach
could be caused by age (more than 60), male (four to five times
(abdominal aorta). In the region around the stomach will be
greater than females), family history (first degree relatives such
much more branching, including to the liver, intestines and
as father or brother), genetic factors, hyperlipidemia (elevated
kidneys. And last, the branching will be forwarded to the
fats in the blood), hypertension (high blood pressure), smoking
direction of human legs (iliac arteries).
and diabetes.
Human blood will be pumped by the heart into the aorta,
which then flows through the artery and its ramifications to all

14 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
Asymptomatic aneurysms may not require surgical According to the result of clinical data, there are difference
intervention until they reach a certain size or are noted to be characteristics based on status of each patient (smoking/ex
increasing in size over a certain period of time. The parameters smoking, fat in blood (dyslipidemied), and hypertency) as
for surgical decisions, but are not limited to, are as follows shown in Table I.
[1,2]:
• aneurysm size greater than 5 centimeters (about two
inches)
• aneurysm growth rate is arround 0.5 centimeters
(slightly less than one-fourth inch) over a period of six
months to one year
• patient’s ability to tolerate the procedure

II. TROMBUS SIGNAL
Thrombosis term will refer to the formation of a blood clot
(thrombus) in the blood vessels or human heart cavities.
Abdominal Aortic Aneurysms are often associated with the (a)
thrombus (clots). This field have been studied and
demonstrated by the pathological, surgical, and clinical
examination based on the results of computed tomography
(CT), ultrasound imaging, angiography, traditional spin-echo
(SE) or cine-MRI. There are many methods have been created
or modified to prove the existence of intact thrombus signal in
the aorta. But until now, with a disorder that occurs in the
aorta, it is difficult to detect or properly evaluate the existence
of thrombus signal [2, 3].

(b)

Figure 3. (a) T1- image and (b) T2- image at the level of Abdominal Aortic
Aneurysms

TABLE I. PATIENT’S CHARACTERISTICS

Figure 2. Aneurysms with a formation of Thrombus [4] Age
Name of Patient Sex Characteristics
(year)
The selection of images for thrombus formation analyzing Patient 1 Male 65 Smooking
is very important. Images are selected from the result of
examination during relaxation took place (as shown in Figure 3 Patient 2 Female 68 Dyslipidémie
of T1 and T2 images)[5]. Smooking, Hypertensi,
Patient 3 Male 62
Dyslipidémie
This work analysed the T1 and T2 of thrombus of SAAA
patient examination to determine whether the thrombus signal Patient 4 Male 82 Ex Smooking
has correlation with the aortic diameter enlargement, and to Patient 5 Male 83 -
predict the rupture risk of the aorta wall.
Patient 6 Male 59 Ex Smooking

III. MATERIALS AND METHODS Patient 7 Male 53 -
Ex Smooking,
A. Data Patient 8 Male 79 Hypertensi,
Data were obtained from 16 patients with Small Abdominal Dyslipidémie
Aortic Aneurysms (SAAA) who have been examined since Ex Smooking,
July 2006 until January 2010. Each patient has been examined Patient 9 Male 77 Hypertensi,
at least 1 to 4 times with examination period between 6 to 12 Dyslipidémie
months (depend on the patient). MR Images were acquired on a Patient 10 Male 71 Smooking,
3T Imager (Trio TIM, Siemens Medical Solution, Germany).

15 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
Age
Name of Patient Sex Characteristics
(year)
Dyslipidémie
Patient 11 Female 74 Ex Smooking
Patient 12 Male 69 -
Ex Smooking,
Patient 13 Male 55 Hypertensi, (a) (b)
Dyslipidémie
Ex Smooking,
Patient 14 Male 51
Dyslipidémie
Ex Smooking,
Patient 15 Male 73 Hypertensi,
Dyslipidémie
Patient 16 Male 59 Smooking
(c)

Figure 5. (a) Anterior-Posterior Diameter, (b) Transversal Diameter, (c)
B. Protocol Small Abdominal Aortic Aneurysms Maximum Diameter
In this study protocol, images originating from: the study of
anatomy, cine-MR images for 3D/4D modeling, images T1/T2,
blood flow images, and images after injection of contrast
agents have been used to study the aspects of inflammation.
For each patient, the images are located in the same position
between one to another examination.

C. Processing
We used MatLab software to precess the data. Preliminary
examination has been conducted for predictive aspect, and final (a) (b)
examination has been conducted as well for data which has
more important thrombus, more areas, and more signals. The
borders have been manually traced to define the Aorta Surface
and Luminal Surface, therefore Thrombus Surface = Aorta
Surface – Luminal Surface, (see figure 4).
In aortic wall surface calculation, thrombus is found if the
thrombus surface area is greater than 30% of aortic surface (c)
area. Diameter of aorta is achieved by tracing manually the
aorta surface. There are three kinds of diameter positions: Figure 6. (a) T1-W image and (b) T2-W image after manual tracing, (c)
Normalization area in the muscle
Anterior-Posterior Diameter, Transversal Diameter and
Maximum diameter, as shown in the figure 5.
D. Paramaters
The muscle signal are slightly differences among each
examinations, therefore we normalized the data of muscle area. Maximum aortic diameter was automatically obtained from
manual tracing on T1 image in all examinations. Then we
calculated the evolution of the aortic diameter (mm/year) = ∆
maximum diameter (mm) / ∆ examination date (day) * 365
days. Several parameters were used to study the thrombus
signal, such as mean, median, standard deviation, skewness that
describes the degree of asymmetry of the signal histogram by
using the equation ∑ni(xi-x)3/Ns3, and the kurtosis that
describes how sharp the peak of the signal histogram which is
defined by using the equation ∑ni(xi-x)4/Ns4-3, where ni is
number of pixel at aorta xi , x is mean value of the aorta, s is the
(a) (b) SD, and N is the total number of pixels. [5]
Each parameter is calculated for the thrombus area, and the
Figure 4. (a) Manual tracing in Aorta Surface, (b) Manual tracing in Luminal signal in the muscle is used to normalize the mean of signal in
Surface (in green line)
thrombus, the median of signal in thrombus and the standard
deviation of signal in thrombus. These parameters are
compared to the evolution of the aortic diameter. By using

16 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
mean/median/SD signal of the aorta and normalized Patient Thrombus Categories Thrombus Categories
mean/median/SD signal of muscle, the thrombus is classified as (based on parameters) (based on visualization)
follow: Homogeneous Thrombus (if T1 = T2 = Low signal); Patient 5 Heterogeneous Heterogeneous
Patient 6 Indefinite Indefinite
Heterogeneous Thrombus (if T1 = T2 = High signal); and Patient 7 Heterogeneous Heterogeneous
Indefinite Thrombus (if T1 ≠ T2 (low and high signal, or high Patient 8 Homogeneous Heterogeneous
and low signal)). Patient 9 Heterogeneous Heterogeneous
Patient 10 Heterogeneous Heterogeneous
IV. RESULTS AND DISCUSSION Patient 11 Homogeneous Heterogeneous
Patient 12 Homogeneous Heterogeneous
We found 13 out of the 16 patients with SAAA have a Patient 13 Homogeneous Homogeneous
thrombus. Figure 7 and 8 are a sample of T1 image which can Patient 14 Heterogeneous Heterogeneous
describe about presence of thrombus in SAAA. Patient 15 Heterogeneous Heterogeneous
Patient 16 Homogeneous Homogeneous

Figure 7. Surface thrombus : 243mm² (11,6%) without thrombus (a)

(b)

Figure 9. P13, Male, 55, ex smooking, hypertensi, dyslipidémie, ∆ Max
Diameter = 2.80 mm/year, 40% surface thrombus, Homogeneous T1 = T2 =
Figure 8. Surface thrombus : 1026mm² (48,4%) with thrombus
Low, T1= 0.391 < 0.815, (b) T2= 0.327 < 0.788

Based on height’s distribution of thrombus signal, there
were 3 patient without thrombus, 5 patiens with homogenous
thrombus, 7 with heterogeneous thrombus and 1 with indefinite
thrombus. Figure 9, 10, and 11 shows the categories of
thrombus presence. If we compare the used of parameters to
the visual, there were 3 differences of the result of thrombus
categories as shown in Table II. There are three categories are
different (patient number 8, 11, 12). It indicates that 81,25% of
thrombus categories determination using parameters are the
same with the result of based on visualization.
(a)
TABLE II. COMPARISON WITH VISUALIZATION CATEGORIES
Patient Thrombus Categories Thrombus Categories
(based on parameters) (based on visualization)
Patient 1 Without thrombus Without thrombus
Patient 2 Without thrombus Without thrombus
Patient 3 Homogeneous Homogeneous
Patient 4 Without thrombus Without thrombus

17 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
and Table IV). From those tables, there are many values of R <
0.3 (not good correlation), a few values of R > 0.3, which
indicates a correlation between thrombus signals and the
evolution of the aortic diameter.

TABLE III. PARAMETERS VS MAXIMUM DIAMETER

Name of T1 T2
Comparison
(b) R² R Equation R² R Equation
Mean/Mean 0.099 0.314 y = 0.030x – 0.010 0.098 y = 0.005x +
Figure 10. P5, Male, 83, ∆ Max Diameter = 2.27 mm/year, 90.85% surface Muscle 0.325 0.485
thrombus, Heterogeneous T1 = T2 = High, (a) T1= 2.675 > 0.815, (b) T2 = Mean/Median 0.099 0.314 y = 0.030x – 0.045 0.212 y = 0.012x +
0.881> 0.788 Muscle 0.329 0.253
Mean/SD 0.006 0.078 y = 0.041x + 0.071 0.266 y = 0.145x +
Muscle 4.002 6.542

Median/Mean 0.101 0.318 y = 0.031x – 0.009 0.097 y = 0.005x +
Muscle 0.364 0.471
Median/Median 0.101 0.318 y = 0.031x – 0.045 0.212 y = 0.011x +
Muscle 0.368 0.249
Median/SD 0.008 0.089 y = 0.045x + 0.063 0.252 y = 0.136x +
Muscle 3.774 6.483

SD/Mean 0.055 0,234 y = 0.005x + 0.004 0.063 y = -0.002x +
(a) Muscle 0.038 0.340
SD/Median 0.055 0,234 y = 0.005x + 0.005 0.069 y = 0.005x +
Muscle 0.0383 0.176
SD/SD Muscle 0.000 0.001 y = -0.000x + 0.000 0.005 y = 0.001x +
1.550 4.873

Skewness 0.093 0.304 y = -0.011x + 0.015 0.121 y = -0.007x +
0.549 0.669
Kurtosis 0.000 0.010 y = -0.000x + 0.040 0.200 y = -0.025x +
2.321 3.649

TABLE IV. PARAMETERS VS ∆ MAXIMUM DIAMETER

(b)
Name of T1 T2
Comparison
Figure 11. P6, Male, 59, ex smoking, ∆ Max Diameter =1.33 mm/year, 6.93% R² R Equation R² R Equation
surface thrombus, Indefinite T1 = Low ≠ T2 = High, (a) T1= 0.691 < 0.815,
(b) T2 = 0.853 > 0.788 Mean/Mean 0.262 0.512 y = 0.024x + 0.029 0.171 y = -0.004x +
Muscle 0.931 0.717
Then all parameters for generate thrombus categories that have Mean/Median 0.001 0.028 y = -0.010x + 0.019 0.134 y = -0.024x +
Muscle 1.125 0.843
correlation with the evolution of aortic diameter compared
with parameters that don’t have correlation with the evolution Mean/SD 0.031 0.176 y = -0.101x + 0.014 0.112 y = -0.160x +
Muscle 5.441 12.953
of aortic diameter which is indicated by many occurrence of
value r <0.3 (r is the coefficient of determination on the
graph). But there are also some parameters indicate a linear Median/Mean 0.000 0.02 y = -0.007x + 0.019 0.137 y = -0.018x +
Muscle 1.109 0.705
correlation between thrombus signal with a maximum
Median/Median 0.001 0.022 y = -0.008x + 0.019 0.137 y = -0.022x +
diameter, where the mean value (r = 0.314), median (r = Muscle 1.123 0.798
0.318), skewness (r = 0.304), or thrombus signal with the Median/SD 0.024 0.154 y = -0.091x + 0.013 0.114 y = -0.156x +
evolution of maximum diameter (mean (r = 0.512). Muscle 5.391 12.437

But there are some parameters that showed a linear
correlation between thrombus signal with a maximum diameter SD/Mean 0.002 0.044 y = 0.003x + 0.022 0.148 y = -0.011x +
Muscle 0.250 0.327
(mean (r = 0.314), median (r = 0.318), skewness (r = 0.304)) or
SD/Median 0.002 0.04 y = 0.002x + 0.027 0.163 Y = -0.013x +
correlation between thrombus signal and the evolution of Muscle 0.252 0.361
maximum diameter (mean (r = 0.512) (as shown in Table III

18 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
SD/SD Muscle 0.005 0.07 y = 0.010x + 0.010 0.102 y = -0.060x + REFERENCES
1.266 5.277
[1] Health Library, Aneurysm Overview, New York-Presbyterian Hospital,
30 Novembre 2008
Skewness 0.046 0.215 y = -0.021x + 0.000 0.017 y = 0.003x + [2] Shin Ito, Koichi Akutsu, Yuiichi Tamori, Shingo Sakamoto, Tsuyoshi
0.178 0.479 Yoshimuta, Hideki Hashimoto, Satoshi Takeshita, Differences in
Atherosclerotic Proﬁles Between Patients With Thoracic and Abdominal
Kurtosis 0.021 0.146 y = -0.016x + 2.000 0.005 y = 0.002x + Aortic Aneurysms, The American Journal of Cardiology, Volume 101,
2.367 2.855 Issue 5, Pages 696-699, 1 March 2008
[3] Toshio Honda, Mareomi Hamada, Yuji Matsumoto, Hiroshi Matsuoka,
Kunio Hiwada, Diagnosis Thrombus and Blood Flow in Aortic
Aneurysms by use of Tagging Cine MRI, International Journal of
V. CONCLUSION Angiology, 6:203-206, 1997
[4] M. Xavier, A. Lalande, P.M.Walker, C. Boichot, A. Cochet, O. Bouchot,
The similarity level between thrombus catagory and E. Steinmetz, L. Legrand, F. Brunotte, Dynamic 4D Blood Flow
visualization is 81.25%. There are no good correlation between Representation in the Aorta and Analysis From Cine-MRI in Patients,
thrombus signal and evolusi of aortic diameter on SAAA Computers in Cardiology, 34:375−378, 2007
(many values of r < 0.3), therefore thrombus signal itself [5] Marco Castrucci et al, Mural Thrombi in Abdominal Aortic Aneurysms:
cannot be used as parameter to predict the evolution of aortic MR Imaging Characterization-Useful before Endovascular Treatment ?,
diameter. The correlation between blood flow, thrombus signal RSNA, 197, Italy, October 1995
and bilogy is still studied. For the next research, we will [6] Christopher M. Kramer, Magnetic Resonance Imaging Identifies the
Fibrous Cap in Atherosclerotic Abdominal Aortic Aneurysm ,
implement other comparation parameters to aortic diameter, Circulation ; 109 ; 1016-1021, 2004
such as: blood flow speed with 3D/4D modeling (The aspect of [7] Eric M. Isselbacher, Thoracic and Abdominal Aortic Aneurysms ,
laminar and turbulance, maximum spped, radial speed, and Circulation, 2005
shear stress). and othehr parameters are comparison our [8] Michèle Coutard, Thrombus versus Wall Biological Activities in
thrombus categories with the visualization categories is 81%. Experimental Aortic Aneurysms, Journal of Vascular Research,2009
For the evolution of the aortic diameter, we found no [9] Shin Matsuoka, Quantification of Thin-Section CT Lung Attenuation in
correlation between thrombus signals with the evolution of the Acute Pulmonary Embolism : Correlations with Arterial Blood Gas
aortic diameter in Small AAA (R < 0.3), but the parameters Levels and CT Angiography, American Roentgen Ray Society,
186:1272-1279, May 2006
were used. The methodologies to measure and the
normalization area with muscle signal will be discussed. We
cannot use thrombus signal alone as a parameter to predict the AUTHORS PROFILE
evolution of the aortic diameter. Relationship between flow
data, thrombus signal and biology findings will be studied. A. Suhendra is a Lecturer of Informatics Engineering, Industrial Engineering
Fakulty of Industrial engineering, Gunadarma university.
Currently, comparison of the blood flow velocity with C. M. Karyati, Graduate from Master Program in Information System,
3D/4D modeling (aspect laminar flow and turbulence, Gunadarma Unviversity, 1998. She is now a Ph.D-Student at Groupe
maximum speed, radial speed, and shear stress) with evolution Imagerie Médicale, Le2i, UMR CNRS 5158, Faculté de Médecine,
Université de Bourgogne, Dijon, France
of the maximum diameter was performed.
A Muslim, Graduate from Master Program in Information System,
Gunadarma Unviversity, 1997. He is a Ph.D-Student at Groupe Database
ACKNOWLEDGEMENTS Sistem Information et Image, Le2i, UMR CNRS 5158, Faculté de
Science et L’Enginer, Université de Bourgogne, Dijon, France
This research was conducted because of aid from the MRI and
A. B. Mutiara is a Professor of Computer Science at Faculty of Computer
Nuclear Medicine Department at the Centre Hospital Science and Information Technology, Gunadarma University
Universitaire (CHU) de Bocage in Dijon, France. More
specifically the authors want to thank Nicolas Abello who has
been so helpful in terms of procurement data. A.B.M. also
gratefully acknowledges financial support of the Gunadarma
Education Foundation during the research.

19 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Empirical Evaluation of the Shaped Variable Bit Rate Algorithm for
Video Transmission
A. Suki M. Arif, Suhaidi Hassan, Osman Ghazali, Mohammed M. Kadhum
InterNetWorks Research Group, UUM College of Arts and Sciences,
Universiti Utara Malaysia,
06010 UUM Sintok, Kedah, Malaysia
{suki1207, suhaidi, osman, kadhum}@uum.edu.my

Abstract—Due to the surge of media traffic over the existing best- video coding, it leaves the flexibility to designers to develop
effort Internet, the network congestion condition is projected to the suitable scheme for specific applications [6]. Thus, any
worsen. Hence, video transmission rate needs to be regulated to created algorithm can be applied to many video coding
adjust it with the network condition or constraint. Therefore, standards.
rate control is essential in video transmission for an accepted
visual quality under some certain rate constraint. One of the Most of the video applications are employing video rate
novel algorithms for the rate control, which is called Shaped VBR controllers in the form of either a Variable Bit Rate (VBR) and
(SVBR), was created by Hamdi et al. SVBR is a novel video data Constant Bit Rate (CBR). The advantage of VBR is that it
rate shaping for a real-time video transmission application. It is a produces a consistent visual quality, while CBR generates
preventive traffic control which allows VBR coding video traffic constant data rate for the network interface.
direct into the network, while regulating unpredictable large
bursty traffic by utilizing a leaky bucket algorithm. SVBR However, the bursty form of VBR causes grave problems to
algorithm uses prediction in calculating the next Group of networks in terms of significant variation of the network traffic,
Picture (GoP) video data size and in determining the next jitters and delays. Equally, the problems with CBR
appropriate quantization parameter value. This algorithm has implementation are the additional delay due to buffering, and
been utilized by many researchers and implemented in many the visual quality tends to vary according to the video content.
network scenarios. However, despite its novel creation for a real-
time, the analytical empirical evaluation in this paper found some
There is an obvious need for an alternative solution by
obvious weaknesses. The weaknesses which are revealed in this taking advantages of both CBR and VBR. At the same time,
paper are the occurrence of a sudden sharp decrease in the data the alternative solution should eliminate the weaknesses of the
rate, the occurrence of a sudden bucket overflow, the existence of both rate controllers. Therefore, an ideal rate controller requires
a low data rate with a low bucket fullness level, and the a higher and consistent visual quality video, video data rate is
generation of a cyclical negative fluctuation. always within a permissible bandwidth level, and less delay.
The delay can happen either because of the introduction of an
Keywords-component; rate control; shaped VBR; video additional buffer or complex computational algorithm.
transmission
By that, it is useful to encode the video with an open loop
VBR as much as possible, but, at the same time, it needs to
I. INTRODUCTION control traffic admission into the network when the permissible
Recent years have witnessed an explosive growth of the level is exceeded. By doing so, it helps to maintain the
Internet and increasing demand for multimedia information consistent visual quality in VBR and get into reasonable
services. Multimedia based applications via the Internet have compromise with adaptive quality when the network
received tremendous attention. In spite of the growing transmission rate degrades. In addition, it avoids unpredictable
networking capabilities of the modern Internet and the large bursty rate variations, as in VBR. However, it is done
sophisticated techniques used by today’s video coding, without the rigidity and systematic coding delay of CBR coders
transmitting video data over the Internet is still a great or intermediate CBR buffer.
challenging task, as stated in [1]. SVBR was introduced by Hamdi et al. [11] in 1997. The
Rate control is always regarded as an essential element of main idea behind SVBR is to limit the open-loop burst while,
the typical video coding as stated in the following works [2], at the same time, allowing open-loop VBR coding, provided
[3], [4], [5], [6]. The general rate control is illustrated in Figure that they are still within a permitted constraint. To achieve this
1. Its main task is to regulate the coded video bits to meet a objective, SVBR manipulates a leaky bucket algorithm to
suitable target rate. Video coding refers to the process of perform admission control. The leaky bucket used in SVBR
reducing the quantity of data used to represent a sequence of can be considered as an imaginary buffer, thus, no extra delay
video pictures or frames. A number of standards of video is introduced. Moreover, Hamdi et al. assumed that for a fast
coding have been defined, such as MPEG-2, H.263, and moving scene with complex image structure, the scene quality
MPEG-4. Although a rate control is not always a part of video can slightly be reduced, since human eyes do not have enough
coding, but it plays a very important role in producing video time to notice the image details. In addition to that, Hamdi et
data traffic [7], [8], [9], [10]. Since, it is not a fix part of the al. have suggested applying the algorithm at Group of Picture

20 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Figure 1: General rate control architecture

(GoP) granularity, which consequently yields to a less complex They claimed that their algorithm is based on SVBR
algorithm and lower delay. algorithm with several enhancements. Among the enhancement
they made, besides implementing SVBR in Evalvid framework
In this paper, we discuss our extensive analysis of the to become Evalvid-RA, they added a supported network
SVBR algorithm performance in identifying its strengths and feedback systems, and performed some changes in the
weaknesses. The organization of the paper is as follows. We parameters of the SVBR's leaky-bucket algorithm. Instead of
describe the related works on the SVBR algorithm in next using SVBR leaky-bucket equation, as in (1), they
section. Then, in Section III, we comprehensively describe the reformulated the equation as (2),
SVBR algorithm. In Section IV, we present our experiments
setting in order to evaluate the SVBR algorithm. After that, in
Section V we elaborate and analyze extensively the experiment  X (k + 1) = min{b, (max{0,X (k) – r} + R(k))} 
results. We summarize our paper in Section VI.
 X (k + 1) = min{b’, (max{0,X (k) – r’} + R(k))} 
II. RELATED WORKS
SVBR can be regarded as an alternative solution that taking where Rk is video data input, rk is the leak rate, b is the
advantages of both CBR and VBR. At the same time, it bucket size and Xk is a level of the bucket fullness after
eliminates the weaknesses of the both rate controllers. Since the transmitting the data in the time of k. In (2), r’ < r, b’ < b,
introduction of SVBR, it has been utilized by several works. where r’ and b’ are dynamically adjusted based on network
However, these works are still implemented under the feedback. r’=roldi/G+rnew(G−i)/G and
limitations or weaknesses of the SVBR. b’=boldi/G+bnew(G−i)/G. When there is no network feedback
during a GOP, r’=rold=rnew and b’=bold=bnew. Here, r' and b' are
Evalvid-RA's adaptive leaky bucket rate and bucket size used
A. Evalvid-RA and RA-SVBR
during GOP-k, rold and rnew are previous and current network
One of the recent works is performed by Lie and Klaue, as update of rate, G is a GOP size, and i is the time index for the
in [12], [13]. They created the Evalvid-RA, a tool-set for rate network feedback event counted as the position in the active
adaptive video performance evaluation in ns-2. The creation of GOP of size G frames.
Evalvid-RA builds on modifications to the original software of
Klaue et al. [14] and Ke et al. [15]. The main modification to What should be stressed here is that although the changes
EvalVid was that the re-assembly post process program had to made in Evalvid-RA is seemed big, but it still maintained the
take into account that multiple MPEG-4 source files and core of SVBR algorithm. By limiting the changes to SVBR
modification to the ns-2 interface, and the associated VBR rate leaky-bucket parameters, such as r’ < r and b’ < b, all the
controller based on SVBR by Hamdi et al. Thus, the engine of weaknesses in SVBR are inherited.
the video rate adaptation in the Evalvid-RA is come from
SVBR algorithm. B. Other Recent SVBR Related Works
Lie and Klaue claimed that Evalvid-RA is a true rate Talaat et al. [16] investigated the effect of incorporating
adaptive video. Their solution has generated a real rate adaptive TFRC on the peak signal-to-noise ratio PSNR of the
MPEG-4 streaming traffic, using the quantizer parameter for transmitted video over the Internet in a simulated environment.
adjusting the sending rate. Then, a feedback based on the VBR They found that TFRC performance on slow motion videos
rate controller is used at simulation time, which supporting was slightly better than on medium-motion that was better than
TFRC and a proprietary congestion control system named P- that on high-motion videos. In this work, they actually
AQM. By simulating in ns-2 of TFRC and P-AQM, they deployed the Evalvid-RA in their investigation, without
demonstrated that Evalvid-RA capabilities in performing close- making any changes to the core of SVBR algorithm.
to-true rate adaptive codec operation with low complexity.

21 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
Another work can be found in [17], where a power important attribute was the long term TCP-friendliness.
management mechanism for wireless video transmission using Although, this work is dissimilar from others in terms of
the TFRC protocol that takes into account feedback about the implementing the SVBR in a multicast environment instead of
received video quality and tries to intelligently adapt unicast, and make some adjustment in data rate, but it does not
transmitting power accordingly. The purpose of the mechanism make any adjustment to the estimation approach in the SVBR.
is to utilize TFRC feedback and thus achieve a beneficial
balance between the power consumption and the received III. SVBR ALGORITHM AND PRINCIPLES
video quality.
The researchers claimed that they have implemented a A. SVBR Fundamentals
module consists of the logic of the proposed mechanism in the It can be concluded that SVBR principles are revolved
Evalvid-RA environment. The module that implemented the around several principles. First, the algorithm works on GoP
TFRC protocol also was changed so that, they claimed, can granularity, thus, it has low complexity that leads to a lesser
provide information about packet losses to their mechanism. delay [12]. Therefore, it is much simpler than that the one
The mechanism calculates the power needed to improve PSNR, employed in the CBR coder which is working on macroblock-
and then this information was passed to the modified wireless by-macroblock variations [11].
physical layer module that is able to increase or decrease power
according to the mechanism. However, as stated previously, the 1) Controlling Video Data Input: The second principle of
fundamental contributors for the weaknesses in SVBR/Evalvid- SVBR algorithm is that it uses a leaky-bucket algorithm to
RA is as a consequence of the estimation and prediction used in avoid an excessive burst data rate into the network. By using
generating the data rate, the work in [17] did not change the leaky-bucket algorithm, the SVBR mechanism can control
anything on this part. the video data admission into the network. Thus, it avoids data
The other work is done by Bouras et al. in [18], [19], where loss at the network interface. This principle can be written as
they performed a performance evaluation of MPEG-4 video follows,
transmission with their proposed multicast protocols, namely
Adaptive Smooth Simulcast Protocol (ASSP) and Adaptive Rk < rk + (b - Xk) 
Smooth Multicast Protocol (ASMP). The features in their
protocols include adaptive scalability to large sets of receivers, The relationship of all variables are illustrated in Figure 2.
TCP-friendly behavior, high bandwidth utilization, and smooth The (b - Xk) in (3) is the space in the bucket that can still
transmission rates which suitable for multimedia applications. accommodate more video data. Whereas, rk is the data from the
They evaluated the performance of their protocols under an bucket that has been sent into the network interface. Thus, the
integrated simulation environment which extends Evalvid-RA blank space in the bucket is now ((b - Xk) - rk). Therefore, (3)
to the multicast domain with the use of the Real-time Transport restricts the video data input, so that the bucket fullness will not
Protocol (RTP) or Real-Time Transport Control Protocol be overflowed. Consequently, no drops will occur.
(RTCP). Simulations conducted under that environment
In order to realize the above-mentioned principle, the video
combine the measurements of network-centric along with video
bit allocation or Rk should be controlled so that the Xk does not
quality metrics. They claimed that the ―joint‖ evaluation
exceed b. For that purpose, Hamdi et al. in [11] proposed (1).
process provides a better understanding of the benefits and
In the Equation, they defined X(k) as a bucket fullness level at
limitations of any proposed protocol for multimedia data
the beginning of GoP-kth video transmission (before
transmission.
transmitting GoP-kth data). Thus, X(k+1) can be regarded as
They called their tool-set as Multi-Evalvid-RA, which the bucket fullness level after transmitting GoP-kth data.
provides all the necessary tools to perform simulation studies Equation (1) restricts the bucket fullness level into (4).
and assess the video quality by using both network related
metrics along with video quality measurements. This is due to
that Evalvid-RA does not support multicast transmission,
which is necessary for experiments and simulations with the
RTP/RTCP protocols. Therefore, they further extend the
functionality of Evalvid-RA by adding the codes in order to
exploit the sender and receiver RTCP reports.
They used Evalvid-RA in implementing media rate control
based on traffic feedback and takes advantage of RTCP's
Sender Report (SR) and Receiver Reports (RR). They claimed
innovation created is the calculation of smooth transmission
rates, which was performed by receivers and is based on RTCP
reports. In such a way, the oscillations are reduced. Another

Figure 2: Restrict input in the leaky-bucket algorithm

22 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
0 < X(k + 1) < b (4)

Here (4) can be realized since the mathematical expression
max{0, (X(k)-r)} from (1) will produce a positive result
between 0 and (X(k)-r), even if the expression (X(k)-r)
produces negative result. Even though expression (max{0,
(X(k)-r)}+R(k)) might produces a value which is higher than b,
the expression min{b,(max{0, (X(k)-r)}+R(k))} will cap the
result to b only.
2) Allowing VBR Coding when the Network Permits: This
principle can be written as follows, as stated in [11], video
sequence with reasonable activity and duration can be coded at
the normal VBR rate while excessively long and/or active
sequence is ―truncated‖ and their bit rate is reduced to r. This
means that for video sequence where conforms to the traffic
contract, the shaping algorithm behaves like normal VBR. On Figure 3: SVBR shaping principle when bucket empty
the other hand, during overload periods (those where video
sequence does not conform to the traffic contract), the
algorithm aims to bring the rate down to r. During these
periods, image quality may be reduced to that of CBR coding.
However, because network resources are dimensioned based
on the leaky-bucket conformance, this shaping avoids data
loss. Thus, only harmful sequences are shaped.
This principle can be depicted in Figure 3 and Figure 4.
SVBR data rates are shown in the red-dotted lines. Here,
SVBR uses bucket fullness level as a conditional parameter to
address this principle. When the bucket is empty, it indicates a
video sequence with a reasonable activity. Whereas, the full
bucket shows that active video sequence is being processed.
This principle can be easily seen in the region II of Figure
3. When the bucket empty, SVBR uses VBR data rate. Thus,
the higher quality of VBR in Region II is maintained without Figure 4: SVBR shaping principle when bucket full
bucket overflow. On the contrary, the video data rate is
decreased into CBR rate when an excessive active sequence is next QP value. The notation Q(k) is used to represent the
occurred, as illustrated in region II of Figure 4. Here, SVBR quantization parameter value used in a kth GoP. Rest(k+1)
compromises high visual quality since the bucket is already notation will be used to represent the estimation data rate.
full. By following VBR data rate, SVBR might risks of bucket
As described in the Subsection III-A2, that the SVBR data
overflowing.
rate lies between CBR and VBR data rate due to the use of
In case of low VBR data rate (lower than CBR rate), as linear calculation. Based on bucket fullness level, SVBR data
seen in the region I and region III of Figure 4, SVBR applies rate can be either close to CBR or VBR. For that purpose,
normal VBR data rate. This will not make the bucket overflow, SVBR uses (5) and (6).
since the leak rate is greater than the input rate. However, in
if Ropen (k + 1) > r,
similar case but with an empty bucket level, SVBR employs
CBR rate. This will increase the video rate which leads to a  Rest(k + 1) = (1 - x) .Ropen(k + 1) + x.r 
better visual quality without compromizing the risk of bucket
overflowing or underflowing. if Ropen (k + 1) ≤ r,
However, in the actual implementation, the bucket rarely  Rest(k + 1) = (1 - x) .Ropen(k + 1) + x.r 
empty or full completely. Thus, by using linear calculation,
which will be described later, the SVBR will be lying between Here Ropen(k + 1) is actually a GoP-(k + 1)th SVBR data rate
CBR and VBR data rate. estimation. Since SVBR is targeted for real time video
application, the next GoP data is not known in advance. Thus,
B. Determining Bit Rate Allocation: Ropen(k + 1) is needed in order to calculate R(k + 1). For that
SVBR uses estimation and prediction in calculating the purpose, SVBR uses (7) as a prediction or estimation for the
next Group of Picture (GoP) video bit rate allocation, that Ropen(k + 1).
means determining R(k+1). The size of the R(k) is very much
related to the QP. Also, SVBR uses estimation to determine the

23 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
Rk   Qk   allocation), suitable QP value should be determined. To
 Ropenk 1  
calculate the next GoP QP value, which is Q(k+1), SVBR
q
requires R(k), q and x. The relationship of all variables with the
SVBR rate control block diagram is depicted in Figure 5.
It is written like that in order to indicate that SVBR will use
quantization parameter q which is used for the generation of The main idea here is as follows, if the current GoP
VBR data rate as its base rate. q is any suitable value that will produces high bit rates data (higher than r) and the bucket
be fixed by user in the beginning of the operation. fullness level is high as well (more than half of the bucket), a
higher QP value for the next GoP should be generated. Higher
Additionally, x is a simple function to calculate the ratio of QP value means lower data rate, R(k+1), which is exactly what
bucket fullness level. The calculation is shown in (8); SVBR is supposed to produce. When the bucket is nearer to
X (k  1) full level and VBR video sequence is active (high rate), the
 x   next data rate should be lower. Thus, the bucket does not tend
b to full or overflow. The opposite will be happened if the current
As described previously, the purpose of the bucket fullness data rate is low. The QP value will be reduced, which will
function is to determine either it should be closer to CBR (r) or produce higher data rate for the next transmission.
SVBR (Ropen(k + 1)). If the bucket fullness is high (x>0.5) and
Ropen(k + 1) (refer to Region II of Figure 4), the SVBR data rate D. Implications of the Design
should be closer to r. This is reflected well in (5); higher x There are many implications from the original principles of
value, increases r value and comparatively will decrease the SVBR design. The implications are inherited from the way
Ropen(k + 1) value by using expression (1-x). On the other hand, SVBR determines (predicts or estimates) the suitable bit rate
if the bucket fullness is low (x>0.5) and Ropen(k + 1)) ≤ r (refer allocation and the QP value. From this estimation, the QP value
to Region I and III of Figure 4), the SVBR data rate is closer to will be fed into the video encoder (refer to Figure 5) to generate
Ropen(k + 1). This condition is also reflected in (5). All the the real encoded video. The data rate for the encoded video
combinations of bucket fullness level, Ropen(k + 1) is greater or might be very much differ from the earlier data rate estimation,
lower than r, and either SVBR is closer to r or Ropen(k + 1) are when estimating the suitable bit rate allocation (Rest(k+1)).
shown in Table I.
Another big implication is on the encoded video data rate.
From (7), it is clear that the next GoP Ropen is calculated Since the estimated QP value is used to encode the video, and
based on previous R(k) and previous QP values. the next original video data rate is much different from the
previous one, this might create another undesired relationship.

C. Determining QP Value E. SVBR Algorithma Flow
In determining the next QP value, which is vital in The gross or important parts of the SVBR algorithm can be
producing the suitable bit rate allocation for the next GoP data, illustrated in a flow chart form as shown in Figure 6. Following
the following equations are used; is the explanation on the notations, commands or variables
if Ropen (k + 1) > r, used;
• X(0): It is the intialization bucket fullness level at
q  Ropen k  1
 Qk  1    GoP-0, which is set to half of the bucket.
1  x  R k  1  x  r 
open
• For k: to perform looping in order to process all video
sequences from the first GoP until the last GoP.
if Ropen (k + 1) ≤ r, • Transmit data: to transmit each GoP data. The real
implementation in the simulation package is much
q  Ropen k  1 more complex. This is due to the fact that the
 Qk  1   
x  R k  1  1  x  r 
open
simulation package imitate almost 100% similar to the
real network behaviour.
As mentioned earlier, QP value is used by the video
encoder to encode the video data. In the context of SVBR, QP
value produces R(k). To obtain a certain R(k) value (bit rate

TABLE 1. RELATIONSHIP OF BUCKET FULLNESS, ACTIVE SEQUENCE AND
SVBR DATA RATE

x < 0.5 x > 0.5
Ropen(k + 1) > r SVBR is closer SVBR is closer
to Ropen to r
Ropen(k + 1) ≤ r SVBR is closer SVBR is closer
to r to Ropen
Figure 5: Rate control block diagram with SVBR

24 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
and approximation in calculating the next GoP VBR
data.
• Ropen(k+1) > r: is to test whether VBR data is in higher
data rate than r. From that test, SVBR can determine
either the next GoP data rate is closer to Ropen or r.
• Q: is to represent Q(k+1), the quantization parameter
for the next GoP video data.

IV. EXPERIMENTS
This section presents a performance evaluation experiments
of SVBR to analyze its strengths and weaknesses.

A. The Evaluation Approach
A combination of several video clips is used for the purpose
of evaluating SVBR. These clips consist of low and high video
rate, thus, the SVBR performance in several scenarios can be
observed.
The video traffic used in this research is the video traffic
trace. The idea is to employ a real video information (from the
selected video sequence) but without the use of a very big data
in the experiments. Compared to traffic model, trace-traffic is
considered credible as it represents an actual traffic load, as
justified in [20]. The video traffic trace is perfectly fix the
requirement and it has been employed by majority of the video
communication research community, as in [21], [22], [23],
[24].

B. The Video Sequences Used
The video sequence used is a combination of several video
clips taken from http://trace.eas.asu.edu/yuv/index.html. These
include ―news‖, ―bridge_far‖, ―bridge_close‖, ―bus‖, and
―highway‖. They are typical video sequences used in various
video rate control studies. All the video clips and its profiles
are describe in Table II. The combined clip duration is 218.4
seconds, which consists of 7099 frames (6552 frames without
Figure 6: SVBR algorithma flow chart header) and 547 GoPs. All clips used are in a raw YUV 4:2:0,
25 frames per seconds (fps) and in CIF format (352x288). CIF
• X(k+1): after the transmission, the virtual bucket and QCIF format are two commonly used formats in video
fullness will be calculated. transmission-related studies [25], [26].
• e: to represent x, which is the ratio of bucket fullness
level to the bucket size.
• Calculate Ropen(k+1): to calculate the next GoP VBR
data. As described previouly, SVBR uses prediction

TABLE 2. PROFILE OF VIDEO SEQUENCES USED IN THE EXPERIMENTS

Video clips Size (frames) Motion categories Description
News 300 Medium Two news reporter are talking.
Bridge (far) 2101 Low A bridge view from far side, then a small
boat going to that direction with a bird flying
in the sky.
Bridge (close) 2001 Low A bridge view from near side and many
people are walking across the bridge.
Bus 150 Very High A bus moving speedily in a road.
Highway 2000 High A panoramic view from a vehicle speedily
moving in a highway.

25 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
C. Rate Control Experiment Settings compared to VBR. This case is clearly can be seen in Figure 7.
In order to execute the experiments, the following The SVBR data rate is bursted from 6048 Bytes/GoP to 253000
parameters setting are employed; Bytes/GoP, while, the VBR data rate is only increased sharply
from 4032 Bytes/GoP to 30240 Bytes/GoP.
• Bucket size (b) = 60000 Bytes/GoP. This size is equal
to 1Mbps transmission speed. This above scenario can be explained as follows;
• Leak rate (r) = 20000 Bytes/GoP. This rate is average For Gop-201, the k is equal to 201. Then, the bucket
open-loop rate for the video sequences used, as fullness ratio for the next GoP, when applying (8), is 0.1008.
suggested by Hamdi et al. [11]. The QP values for the respective GoPs are shown in the Table
• VBR Initial q = 10. This initial value is selected to III.
provide space for the SVBR rate control to increase or
decrease QP value freely. For a more limited network X 202 6048
x   0.1008
resources setup, the higher initial q value should be b 60000
considered.
R201  Q201 6048  7
Ropen202    4233.6
V. THE RESULT AND ANALYSIS q 10
The strengths and weaknesses of the SVBR algorithm will
Since Ropen(202) is less than r (CBR=20000), then (6) is
be highlighted based on the analysis on several scenarios
applied;
below. The analysis are based on how SVBR will react when
certain scenarios occur. The analyzed scenarios include the q  Ropen202
Q202 
occurrence of sharp decrease in VBR data rate, sudden bucket
overflow, low data rate with low bucket fullness level, and x  R 202  1  x  r 
open

fluctuate data rate.
10  4233.6
Q202 
A. Sharp Decreases in SVBR Data Rate 0.1008  4233.6  1  0.1008  20000
Figure 7 show the phenomenon of sharp decreses in SVBR
data rate which can be seen at the GoP-201. Figure 7 charts the
Q202 
42336 42336
  2.2995
data for this particular GoP and the GoPs around it. The sudden
sharp decrease in SVBR data rate is from 17,136 Bytes/GoP to
426.7469  17984 18410.7469
6,048 Bytes/GoP. The SVBR data rate is examined here
because the generation of the next SVBR data rate is very Since the nearest QP value for Q(202) is 2, then Q(202)=2.
much dependent on it (refer to Figure 6); Rest(k+1) and What should be highlighted here is that, when at GoP-201 the
Ropen(k+1) is the SVBR data rate estimation for the GoP-(k+1). data rate is comparatively very low, the algorithm wants to
increase the data rate for the next GoP by lowering the QP
What can be observed here is that when a sharp decrease in value. The algorithm succeeds in its principle to perform that
SVBR data rate occurs, it will automatically generate a sudden task. The QP value has been decreased from 7 to 2, with the
bursty in the next SVBR data rate. This kind of scenario occurs intention to increase the data rate substantially. However, at
especially when the next VBR data rate is increased sharply as GoP 202, VBR data rate increases to higher rate, consequently
well. The possible problem of this scenario is that, the SVBR QP=2 for SVBR GoP 202 data rate is equal to 252000
data rate might be increased at very much higher rate as Bytes/GoP. For a reference purpose, if VBR data rate remains
at the same rate for GoP-202 as the GoP-201, QP=2 will
moderately increase SVBR GoP 202 data rate to 24192
Bytes/GoP.
From the above calculation, it can be concluded that one of
the weaknesses of SVBR algorithm is when a sharp decrease in
VBR data rate occurs, the SVBR data rate for the next GoP will
increase dramatically. This is especially true in the case the
next VBR data rate is increased sharply as well. The sudden
sharp fluctuation in the SVBR data rate can go up beyond the
permissible level, which overflows the bucket. This scenario
crops up as a result of very low bucket level fullness and low
Ropen value case. This leads to a very low QP value and
consequently, producing very high data rate.

TABLE 3. RESPECTIVE QP VALUES FOR GOP 198-204

GoP # 198 199 200 201 202 203 204
Figure 7: Sharp decrease scenario in the GoP-201
QP 7 7 7 7 2 25 18

26 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
B. When a Sudden Bucket Overflow Occurs fullness is also high, the estimation for the R(370) is going
closer to r or CBR. All these will produce a higher QP value
Figure 8 shows the case when a sudden bucket overflow (Q(370)) than q. Accordingly, the higher QP value will
occurs. This scenario can be observed starts from GoP 369 generate low SVBR data rate. This is a good strentgh of the
until 381. This phenomenon can be described as an advantage SVBR algorithm. Even though the VBR data rate are very high
of the SVBR algorithm, but at the same time, it shows poor and it is maintained high for several consecutive GoPs, SVBR
performance as well. The disadvantage of this algorithm can be algorithm quickly adjust its data rate to an acceptable level and
seen at GoP-369. Here, when VBR data rate fluctuates sharply maintains that rate onwards.
to 59472 Bytes/GoP, exceeds the bucket size, the SVBR also
follows this behaviour, even to higher data rate than the VBR. It can be concluded here that when there is a high data rate
burst occurs, for the next first GoP, SVBR will burst as well
In addition to the aforementioned explanation, since SVBR leading to bucket overflow. As each GoP consists 12 frames,
uses previous GoP data, which is GoP 368, to estimate the next this means that 12 frames will burst as well. However, SVBR
GoP data, it generates a small Ropen. The small Ropen and higher has been designed to quickly retract the burst into permissible
bucket fullness level will generate R(369) (refer (6)) close to level for the following GoPs.
the Ropen value (refer to Table I). Consequently, SVBR
algorithm generates QP value which is close to the default C. Low Data Rate and Low Bucket Fullness Level
value q. This form of estimation is good for video sequences
which are increased or decreased in a smoother way. But, for As shown in Figure 9, this scenario can be clearly seen
the next video sequence, even though the date rate is sharply from GoP 26 to 200. From this range, especially from GoP 106
increased, the q value is still the same as for previous video to 200, it shows a long flat low rate sequence. Part of this
sequence. Then, for the SVBR, when the QP value is around sequence is shown in Figure 9. Here, both of CBR and VBR
the default q value, it produces a high data rate. This is due to are at low rate, SVBR is seen tend to be around the both rates.
that SVBR is using the real data for the transmission, which is SVBR should increase its data rate here to improve its visual
same as VBR (Q(369)=q=10). quality. It is clearly shows here that SVBR is not designed to
increase the visual quality in a low data rate sequence.
X 370 60000
x   1.0000 Based on Figure 9, for SVBR to estimate GoP 106 from
b 60000 GoP 105, SVBR bases on its distance to CBR (r) and VBR. In
the case of calculating GoP 106 using GoP 105, that its data
R369  Q369 59472  10 rate is very close to r, this derives SVBR to adjust its rate to
Ropen 370    59472
q 10 lower than r value. It does that by lowering the R(106)
estimation than the real value of the R(105), refer to (6). By
Since Ropen(370) is greater than r, then (9) is applied; designing in such a way, it produces a higher QP value
q  Ropen 370
estimation, which is 7 (previously QP=6). Consequently, it
Q370  generates lower value of data rate for SVBR GoP 106 data rate.
x  R 370  1  x   r 
open
In producing SVBR data rate for GoP 107, the SVBR will
estimate R(107) as a little bit higher than the real R(106) since
10  59472 R(106) is located in the middle of the r and VBR but at a little
Q370 
594720
  29.736
0  59472  1  20000 20000 bit lower distance. In consequence, when calculating the value
for QP, it gains Q(107)=6.7717. Since, the nearest round
Since the nearest QP value for Q(370) is 30, then QP=30. number for 6.7717 is 7, it produces the SVBR data rate for GoP
The data rate for QP=30 is 25200 Bytes/GoP, then accordingly 107 similar to the previous GoP. This scenario is continuing
SVBR data rate for GoP-369 is 25200 Bytes/GoP. until the GoP 200, where VBR data rate for GoP changes to
other value.
In the above case, although the present VBR and estimation
of the next VBR (Ropen(370)) are high, and since the bucket Therefore, the clear disadvantage of SVBR algorithm is that
it has been designed to maintain its data rate to be vary between
VBR and CBR data rate. Inevitably in a low data rate,

Figure 9: A long-low-flat data rate sequence
Figure 8: Sharp VBR increase scenario

27 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
especially in a long-flat data rate, SVBR tend to be around that
rate. Thus, it maintains its visual quality. However, since the
bucket fullness is in a low level, the data rate can be increased
so that the visual quality can be increased as well.

D. Fluctuate Data Rate
There are two types of conflicting fluctuated data rate;
positive and negative fluctuation. Video data rate fluctuates
generaly. A positive fluctuation is a condition where its QP
Figure 10: SVBR data rate fluctuation
values are constant. Thus, video data rate fluctuation is
considered as a positive fluctuation. A negative fluctuation is a
condition where data rate and QP value are not constant. This
subsection is referred to the negative fluctuation, especially the
fluctuation in the SVBR data rate. In the following, two cases
of the negative fluctuation are explained.
Case I: VBR is Higher than CBR
Wherever SVBR fluctuates differently from VBR, it can be
considered as a negative fluctuation. One of this scenario can
be seen at GoP 208 to 250. Figure 10 and 11 show this
scenario. Figure 11: SVBR QP fluctuation

This scenario is one of the good examples of the sensitive The fluctuation occurs also as a result of sensitivity in the
relationship as describe in the subsection III-D. In the GoP 209, algorithm when the gap between VBR and CBR is quite wide.
QP=11 and SVBR data rate (R(209)) is 19152. This data rate is Although, it is less wider as compared to the previous
below both CBR and VBR data rate at that GoP. One of the subsection, but the changes in one QP value has contributed to
principles of SVBR is to let the next data rate lying between this fluctuation. Another attributes to this fluctuation is the
VBR and CBR data rate by adjusting the next QP value. It oscillation on the bucket fullness level at the CBR data rate.
should be emphasized here that SVBR is only able to estimate
the next QP value with the hope that the actual data rate will be
lying between intended region. VI. SUMMARY
This paper presented the extensive analysis of the SVBR
Since the generated Ropen for GoP 210 is a little bit higher, it
algorithm performance. In general, this algorithm has
will produce a little bit higher QP value but below than 11.
demonstrated its novel creation of an ideal rate controller
Another way to comprehend this scenario is by looking at how
especially for real time video application which produce a
SVBR try to increase its data rate to be between CBR and
higher visual quality video, the video data rate is within a
VBR. Given that the current data rate is below CBR-VBR and
permissible bandwidth level, and less delay as a result of no
its QP value is 11, SVBR will decrease its QP value (to
introduction of an additional buffer or complex computational
increase its data rate) but to the maximum of 10. This is due to
algorithm.
that the QP value for the VBR is 10. The calculation of Q(210)
gives 10.4223 and since there is no fraction of QP encoding, However, there are spaces for performance improvement of
SVBR obtains Q(210)=10. the SVBR algorithm, especially under certain specific
The Q(210)=10 will generate similar data rate with VBR
GoP-210 data rate. Since, VBR data rate is in active sequence,
SVBR obtains much higher data rate compared to previous data
rate. The negative implication of this situation is on the
generation of the next QP and its data rate. As a result of the
wider gap between Ropen and r, it causes the generation of QP
becomes sensitive. In the case of Q(211), SVBR obtains
14.3137 (or 14 only). This QP, when translates to the actual
data rate, is equal to 17136 Bytes/GoP. It creates a significant
ding-dong in the SVBR data rate. This situation continues until Figure 12: SVBR with VBR lower than CBR data rate fluctuation
GoP 250 when VBR data rate changes into a lesser active
scene.
Case II: CBR is Higher than VBR
At GoP 424 to GoP 448, similar negative fluctuation can be
observed but this time VBR is lower than CBR. The other
dissimilar condition is that the bucket fullness level is also
fluctuating between above and below CBR data rate. Figure 12 Figure 13: SVBR with VBR lower than CBR QP fluctuation
and 13 show this scenario.

28 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
conditions. Besides the strengths of SVBR algorithm, this [16] M. A. Talaat, M. A. Koutb, and H. S. Sorour, ―PSNR Evaluation of
analysis had pointed several weaknesses areas that can be Media Traffic over TFRC,‖ International Journal of Computer Networks
& Communications (IJCNC), vol. 1, no. 3, pp. 71–76, 2009.
considered for improvement. The circumstances that had been
[17] C. Bouras, V. Papapanagiotou, K. Stamos, and G. Zaoudis, ―Efficient
strongly stressed were when there is an occurrence of a sudden power management adaptation for video transmission over TFRC,‖ in
sharp decrease in the VBR data rate, the occurrence of a sudden 2010 Sixth Advanced International Conference on Telecommunications
bucket overflow, the existence of a low data rate with low (AICT), May 2010, pp. 509 –514.
bucket fullness level, and the generation of a cyclical negative [18] C. Bouras, A. Gkamas, and G. Kioumourtzis, ―Adaptive Smooth
fluctuation. These weaknesses play an important role in Simulcast Protocol (ASSP) for video applications: Description and
designing a new algorithm for stored video transmission performance evaluation,‖ Journal of Network and Systems Management,
pp. 1–35, 2010, 10.1007/s10922-010-9159-8. [Online]. Available:
application. http://dx.doi.org/10.1007/s10922-010-9159-8
[19] ——, ―Evaluation of single rate multicast congestion control schemes
REFERENCES for MPEG-4 video transmission,‖ in Proceedings of the 5th Euro-NGI
conference on Next Generation Internet networks, ser. NGI´09.
[1] P. Seeling, M. Reisslein, and B. Kulapala, ―Network performance Piscataway, NJ, USA: IEEE Press, 2009, pp. 32–39. [Online]. Available:
evaluation using frame size and quality traces of single-layer and two- http://portal.acm.org/citation.cfm?id=1671421.1671426
layer video: A tutorial,‖ IEEE Communications Surveys Tutorials, vol.
6, no. 3, pp. 58 –78, 2004. [20] A. Al Tamimi, R. Jain, and C. So-In, ―Modeling and prediction of high
definition video traffic: A real-world case study,‖ in 2010 Second
[2] M.-J. Kim and M.-C. Hong, ―Adaptive rate control scheme for real-time International Conferences on Advances in Multimedia (MMEDIA), Jun.
H.264/AVC video coding,‖ in 2010 Digest of Technical Papers 2010, pp. 168 –173.
International Conference on Consumer Electronics (ICCE), 2010, pp.
271–272. [21] W.-K. Liao and Y.-H. Lai, ―Type-aware error control for robust
interactive video services over time-varying wireless channels,‖ IEEE
[3] M. Semsarzadeh, M. Langroodi, and M. Hashemi, ―An adaptive rate Transactions on Mobile Computing, vol. 10, no. 1, pp. 136 –145, Jan.
control for faster bitrate shaping in x264 based video conferencing,‖ in 2011.
2010 IEEE International Symposium on Broadband Multimedia Systems
and Broadcasting (BMSB), Mar. 2010, pp. 1 –4. [22] R. Zhang, R. Ruby, J. Pan, L. Cai, and X. Shen, ―A hybrid
reservation/contention-based MAC for video streaming over wireless
[4] X. Liu, Q. Dai, and C. Fu, ―Improved methods for initializing R-Q networks,‖ IEEE Journal on Selected Areas in Communications,, vol.
model parameters and quantization parameter in H.264 rate control,‖ in 28, no. 3, pp. 389 –398, Apr. 2010.
2009 WRI World Congress on Computer Science and Information
Engineering, vol. 6, Mar. 2009, pp. 320 –323. [23] M. S. Koul, ―Error Concealment and Performance Evaluation of
H.264/AVC Video Streams in a Lossy Wireless Environment,‖ Dept. of
[5] S. G. Zhao Min, Takeshi Ikenaga, ―A novel rate control algorithm for Electrical Engineering, The University of Texas at Arlington, Tech.
H.264/AVC,‖ in The 23rd International Technical Conference on Rep., 2008.
Circuits/Systems, Computers and Communications. The Institute of
Electronics, Information and Communication Engineers (IEICE), 2008, [24] O. A. Lotfallah, M. Reisslein, and S. Panchanathan, ―A framework for
pp. 725 – 728. advanced video traces: evaluating visual quality for video transmission
over lossy networks,‖ EURASIP J. Appl. Signal Process., pp. 263–263,
[6] Z. Chen and K. N. Ngan, ―Recent advances in rate control for video Jan. 2006. [Online]. Available:
coding,‖ Image Commun., vol. 22, no. 1, pp. 19–38, 2007. http://dx.doi.org/10.1155/ASP/2006/42083
[7] H. A. Marios C. Angelides, The Handbook of MPEG Applications: [25] C.-H. Ke, C.-K. Shieh, W.-S. Hwang, and A. Ziviani, ―Improving video
Standards in Practice. John Wiley and Sons, 2011. transmission on the Internet,‖ IEEE Potentials, vol. 26, no. 1, pp. 16–19,
[8] S. Eshaghi and H. Farsi, ―Rate control and mode decision jointly Jan. 2007.
optimization in H.264AVC,‖ in Proceedings of the 4th conference on [26] P. Seeling, F. H. P. Fitzek, and M. Reisslein, Video Traces for Network
European computing, ser. ECC’10. Stevens Point, Wisconsin, USA: Performance Evaluation: A Comprehensive Overview and Guide on
World Scientific and Engineering Academy and Society (WSEAS), Video Traces and Their Utilization in Networking Research. Springer,
2010, pp. 280–283. [Online]. Available: 2007.
http://portal.acm.org/citation.cfm?id=1844367.1844414.
[9] P. Ranganth and Y. V. Swaroop, ―Adaptive rate control for H.264 video
encoding in a video transmission system,‖ in XXXII National Systems AUTHORS PROFILE
Conference (NSC 2008), Dec. 2008. Ahmad Suki Che Mohamed Arif is is a lecturer in the Graduate Department
[10] W. Lu, X. Gao, Q. Deng, and T. Wang, ―A basic-unit size based of Computer Science, Universiti Utara Malaysia (UUM) and is currently
adaptive rate control algorithm,‖ in Fourth International Conference on attached to the InterNetWorks Research Group at the UUM College of Arts
Image and Graphics (ICIG 2007), Aug. 2007, pp. 268 –273. and Sciences as a doctoral researcher.
[11] H. Hamdi, J. W. Roberts, and P. Rolin, ―Rate control for VBR video Associate Professor Dr. Suhaidi Hassan is currently the Assistant Vice
coders in broad-band networks,‖ IEEE Journal on Selected Areas in Chancellor of the College of Arts and Sciences, UUM. He is a senior member
Communications, vol. 15, no. 6, pp. 1040–1051, 1997. of the Institute of Electrical and Electronic Engineers (IEEE) in which he
actively involved in both the IEEE Communications and IEEE Computer
[12] A. Lie and J. Klaue, ―Evalvid-RA: Trace driven simulation of rate
societies. He has served as the Vice Chair (2003-2007) of the IEEE Malaysia
adaptive MPEG-4 VBR video,‖ Multimedia Systems, vol. 12, no. 1,
2008. Computer Society.
Dr. Osman Ghazali is a Senior Lecturer at Universiti Utara Malaysia. As an
[13] A. Lie, ―Enhancing rate adaptive IP streaming media performance with
academician, his research interests include congestion control, quality of
the use of Active Queue Management,‖ Ph.D. dissertation, Norwegian
University of Science and Technology, 2008. services, wired and wireless network, transport layered protocols and network
layered protocols. His works have been published in international conferences,
[14] J. Klaue, B. Rathke, and A. Wolisz, ―Evalvid - a framework for video journals and won awards on research and innovation competition in national
transmission and quality evaluation,‖ in the 13th International and international level.
Conference on Modelling Techniques and Tools for Computer
Performance Evaluation, 2003, pp. 255–272. Dr. Mohammed M. Kadhum, is a visiting assistant professor in the Graduate
Department of Computer Science, UUM. He is currently attached to the
[15] C.-H. Ke, C.-K. Shieh, W.-S. Hwang, and A. Ziviani, ―An evaluation InterNetWorks Research Group at the UUM College of Arts and Sciences as a
framework for more realistic simulations of MPEG video transmission,‖ research advisor. He has been awarded with several medals for his outstanding
Journal of information science and engineering, vol. 24, pp. 425–440, research projects.
2008.

29 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

An Efficient Self-Organized Authentication and Key
Management Scheme for Distributed Multihop Relay-
Based IEEE 802.16 Networks
Adnan Shahid Khan, Norsheila Fisal, Sharifah M. Abbas
Kamilah, Sharifah Hafizah, Mazlina Esa, Wireless Communication Cluster
Zurkarmawan Abu Bakar MIMOS Berhad, Technology Park Malaysia
UTM-MIMOS Center of Excellence in 57000 Kuala Lumpur, Malaysia
Telecommunication Technology, Faculty of Electrical mazlan.abbas@mimos.my
Engineering, Universiti Teknologi Malaysia 81310
Skudai, Johor, Malaysia, adnan.ucit@gmail.com,
{sheila,kamilah,sharifah, mazlina,
zurkarmawan}@fke.utm.my,

Abstract— Wireless internet services are rapidly expanding and between an Multihop Relay Base Station (MR-BS) and an
improving, it is important to provide users with not only high Mobile Station (MS), here Relay Station (RS) is just an
speed and high quality wireless service but also secured. amplify and forward, but in the second security mode, referred
Multihop relay-based support was added, which not only help for to as distributed modes, which incorporate authentication and
improving coverage and throughput but also provides features key management between an MR-BS and a non-transparent
such as lower backhaul deployment cost, easy setup, robustness RS we called as NRS and between the NRS and a MS. During
and re-configurability, which make it one of the indispensable the registration process, an RS can be configured to operate in
technologies in next generation wireless network. A WiMAX distributed security mode based on its capability [1]. Since
network usually operates in a highly dynamic and open AUTH-INFO message is optional and informative we begin
environment therefore it is known to be more vulnerable to with the security analysis from the AUTH-REQ message. As
security holes. Security holes most of the time is trade off with
this message is plain text and for such message, eavesdropping
authentication and key management overheads. In order to
operate securely, communication must be scheduled either by a
is not a problem since the information is almost public and is
distributed, centralized or hybrid security control algorithms preferred to be sent in plain text to facilitate authentication. To
with less authentication and key management overheads. In this capture and save the authentication message sent by a
paper, we propose a new fully self-organized efficient legitimate, is not a big deals, thus NRS may face a replay
authentication and key management scheme (SEAKS) for hop- attack from an adversary. Although an adversary
by-hop distributed and localized security control for Multihop eavesdropping the message, cannot derive the AK from the
non-transparent relay based IEEE 802.16 networks which not message, because it does not have the corresponding private
only helps in security counter measures but also reduce the key. However, the adversary still can replay message II
authentication and key maintenance overheads. The proposed multiple times and then either exhaust NRS capabilities or
scheme provides hybrid security controls between distributed force NRS to deny the SS who owns that certificates [1] [2].
authentication and localized re-authentication and key The reason is that if NRS sets a timeout value which makes
maintenance. The proposed scheme uses distributed non- NRS reject Auth REQ from the same MS in a certain period ,
transparent decode and forward relays for distributed the legitimate request from the victim MS will be ignored.
authentication when any non-transparent Relays (NRS) want to Then denial of service attack occurs to victim MS, however
join the networks and uses localized authentication when NRSs the ultimate solution for these types of attacks are the
want to re-authenticate and do key maintenance. We analyze the introduction of digital signatures at the end of the messages
procedures of the proposed scheme in details and examine how it which can be automatically time-stamped, that basically
works significantly to reduce overall authentication overheads
provides the authentication and non-repudation of this
and counter measures for security vulnerabilities such as Denial
of Service, Replay and interleaving attacks.
message. The design of digital signature system may be
flawed or vulnerable to some specific attacks such as collision
attacks against X.509 public-key certificates and
Keywords- Wimax Security, Multihop Relay based IEEE cryptographically weak pseudo random bit generator.
802.16, Key Management, Self-Organized Authentication) Adversaries may attempt for total break, universal forgery,
selective forgery or existential forgery.
I. INTRODUCTION
The strongest security definition requires protection against
In Multihop Relay (MR) network, two different security existential forgery even if an adversary is able to mount an
modes are referred, the first one is referred to as the adaptive chosen message attack. Later, nonce was added to the
centralized security mode which is based on key management

30 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
digital signature, the idea of nonce values is that they are used literature is very sparse. In this network, all the relays are
only once with a given key, however, the exchange of nonce connected to MR-BS wirelessly and transparently or non-
only assures SS that message III is a replay corresponding to transparently and only MR-BS is connected to IP cloud as a
its request. The NRS still faces the replay attack because NRS backhaul, thus this infrastructure can be used in many real
cannot tell whether message II is sent recently or it is just an time applications [2].
old message [3]. If reply attack cannot be successful, for sure As the matter of fact, security is essential in wireless
‘denial of service’ will occur. The author of [4] also suggested technologies to allow rapid adoption and enhance their
passing the pre-AK to SS instead of AK and let SS and NRS maturity. Due to lack of physical boundaries, the whole relay-
derive AK from pre-AK at both ends. If the generation of AK based infrastructure in exposed to security holes. However,
exhibits significant bias, adding freshness in the AK may IEEE 802.16 standard stipulates some powerful security
prevent the exposure of the AK, however according to [4] this controls, including PKMv2, EAP-based authentication and
cannot provide freshness as they claimed. If we consider the over-the-air AES based encryption. But secure technology
security issues of relay-based IEEE 802.16 networks in does not in itself comprise a secure end-to-end network and
centralized as well as distributed authenticated, every node consequently, WiMAX presents a range of security
need to authenticate itself with MR-BS and ultimately with vulnerabilities. Since the first Amendment was released on
AAA server. Secondly, every node needs to maintain two MR specifications [1], a few papers have been published to
simultaneously keys AK and TEK to remain authenticated. introduce and address the security issues. There are some
Failure to maintained these keys will result in the re- papers that review this standard in details such as [6] and [7],
authenticated from scratch which is no doubt extra and there are some papers they purely works on key
managements specially Sen Xu and Manton Mathews who
authenticated overhead. Let’s suppose, there are five NRS,
published a series of work such as [3] and [4] on security
where every NRS has to keep track of its AK and TEKs and issues on the standard as well as on Privacy key Management
consequently authentication. Thus generation of authentication (PKM) protocols. Karen Scarfore with her team came up with
overhead by five NRS no doubt lessen the overall deployed a special publication on Guide to security for Wimax
network efficiency. To solve this authentication overhead technologies (Draft) which was the recommendations of the
problem, Self organized and efficient authentication and key National Institute of Standards and Technology (NIST).
management scheme (SEAKS) proves to be the best candidate Taeshik Shon and Wook Choi [8] discussed about the
in the relay-based IEEE 802.16 network, which utilized non- Analysis of Mobile WiMAX Security, Vulnerabilities and
transparent and decode and forward relays. SEAKS provides Solutions. Y. Lee and H. K. Lee in their paper [9] gives more
hybrid scheme with distributed authentication and localized focus on hybrid authentication scheme and key distribution for
re-authentication and key maintenance. However, this MMR in IEEE 802.16j.
technique not only helps in minimizing the overall
The authors [10] and [11] review the standard and
authentication overhead on MR-BS and AAA server but also analyzed its security in many aspects, such as vulnerabilities in
provide efficient way to countermeasure the vulnerabilities. authentication and key management protocols and failure in
data encryption. In IEEE 802.16j [12] standard, Multihop
The rest of the paper is organized as follows, after related Relay (MR) is an optional deployment in which a BS in
work, section 3 gives the overview of generals attacks on (802.16e) may be replaced by a Multihop Relay BS (MR-BS)
network, section 4 discusses centralized and distributed and one or more relay stations (RS). The MR mechanism
authentication controls, section 5 deals with the security goals provides several advantages, such as providing additional
of relay-based WiMAX network, section 6 describe the self- coverage for the serving BS, increasing transmission speed in
organize scheme (SEAKS), section 7 gives the analysis of an access network, providing mobility without SS handover,
proposed scheme which is followed by conclusion and future decreasing power consumption when transmitting and
work. receiving packets, and enhancing the quality of services [3].
II. RESEARCH BACKGROUND
There has been a significant amount of work done on security
issues and their protocols as shown above but none of these
In 2006, the IEEE 802.16 working group (WG) approved cover security protocols which works for minimized
a project Authorization Request (PAR) focused on the Relay authentication and key management overheads in non-
Tasks Group (TG). The main task of this Relay TG was to transparent Relay-based WiMAX networks in distributed
develop an amendment to the IEEE Std 802.16 enabling the environment.
operation of Relay Station (RSs) in OFDMA wireless
networks defined by 802.16 [2]. Enhancement of Relays to III. GENERAL ATTACKS ON RELAY-BASED IEEE 802.16
support Multihop not only increases the wireless converge but NETWORK
also provide features such as lower backhaul deployment cost, Before we start to elaborate our self organized algorithm,
easy setup and high throughput. Relay stations concept as we would like to high-light some of the typical MAC layer
discussed in [1][2] and [5] introduced four types of RSs from attacks on authentication and key management protocols. The
the perceptive of physical and Mac layer. After successful first and very common attack is message replay attack [7].
comparison, the main focus of this research is on the non- This attack is not only common in key management and
transparent RS operating in distributed scheduling and security authentication protocols but also in multicast and broadcast (M
mode [2], WiMAX relay-based network in still under draft and & B) services [11]. In a replay attack, an adversary intercepts

31 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
captures and saves the authentication messages sent by the have this key information. The intermediate RS use particular
legitimate RS/SS. Thus adversary impersonates the legitimate shared keys to authenticate management messages which
RS/SS and resends this message after specific period of time. received from other RSs [12][14].
Denial of service (DoS) is also one of the major attacks in
wireless networks especially in WiMAX networks. Here, B. Distributed Security Control
consider an adversary that eaves-dropping the message cannot In this mode, an access RS, which provides a point of
derive the AK as it does not have the corresponding private access into the network for an MS or RS, can derive the
key. This adversary still can replay AUTH-REQ message authentication key established between MS and MR-BS. An
multiple times and thus exhaust MR-BS capabilities and force RS can be configured to operate in distributed security mode
MR-BS to deny this adversary. This may happen, if the MR- based on its capability during the registration process, and
BS sets a time out value which makes MR-BS reject AUTH- relays initial key management messages between the MR-BS
REQ message from the same RS/SS with an interval of time. and MS/subordinate RS. Upon master session key
Thus, MR-BS denies the legitimate RS/SS AUTH-REQ, establishment, access RS securely acquires relevant
which actually owns the certificate. DoS are common in Authorization Key of the subordinate RS/MS from the MR-
authentication, key management protocols and M & B BS. Using PKM protocol, the access RS can derives all
services. Man-in-the-Middle (MiTM) attack is another critical necessary keys. Different traffic encryption keys (TEKs) are
attack and is generally applicable in communication protocol used for relay link and access link in distributed security
scheme where mutual authentication is absent especially in control mode. They are distributed by MR-BS and RS
PKMv1. This attacks leads to message modification and respectively [4][15]. The SA will be created between an MS,
masquerading problems, specially node spoofing, rogue base an access RS and the MR-BS in distributed security mode.
as well as relay stations, theft of service (ToS). To avoid Each MS shall establish an exclusive primary SA with the RS,
MiTM attack on PKM protocol, mutual authentication was interacting with the RS as if it were a BS from the MS’s view.
proposed i.e. PKMv2. No doubt PKMv2 is soundly safe for Similarly, each RS shall establish an exclusive primary SA
MiTM but it cannot help allowing adversary to play with MR-BS [12][16].
interleaving attack.
Interleaving attack in complex to be explained but easy to
attempt. An adversary attempts this attack with the help of two V. SECURITY GOALS OF RELAY-BASED WIMAX
different instances. In the first instance, adversary NETWORKS
impersonates as SS/RS and sends the interrupted message to Non-transparent Relay-based WiMAX network may
the MR-BS. MR-BS authenticates and replied with require the following security function, which have not widely
corresponding keys. Adversary needs to reply these keys to been studied by others until now.
RS/SS to be successfully authenticated, as it cannot decrypt
the message encrypted by the SS/RS’s public key in order to • Localized and hop-by-hop authentication is required.
get the AK to encrypt the nonce challenge. Thus, it cannot do In Relay-based WiMAX network. NRS in introduced
authentication currently. Now to solve this technicality, for coverage extension and throughput enhancement,
adversary force RS/SS to run another protocol instance to for this purpose, hop-by-hop authentication between
answer the challenge. Once RS/SS send the request, adversary NRS, NRS/MS and NRS/MR-BS should be
replies SS with the same nonce challenge which the MR-BS supported for self organized network operations.
sends him. Thus RS/SS send nonce and AK to adversary • All the participating devices must be validated and
which later sends to MR-BS to finish this authentication authenticated by AAA server through MR-BS,
successfully. This attack normally can occur only on PKMv2 because digital certificates of participating devices
or where mutual authentication is present. In IEEE 802.16 are only registered in AAA server database, however,
Multihop networks, the number of wireless devices engross is NRS should authenticate other NRS/MS on behalf of
increased, thus produce wide space for interleaving attack [3] MR-BS, and basically this concept leads our
[4]. proposed scheme towards self organized way.
• Conventional MS should be used in non-transparent
Relay-based WiMAX network without any functional
IV. CENTRALIZED VS. DISTRIBUTED AUTHENTICATION modification in MS.
• Overall authentication overhead should be
A. Centralized Security Control minimized.
In this mode, the intermediate RS is not involved with the
In this paper we proposed self organized distributed and
establishment of the security association (SA) between MS
localized authentication and key management, where initially
and MR-BS in the multihop relay system. The RS only simply
participating devices validated and authenticated by MR-BS
relays the user data or MAC management message that it
and afterward NRSs are responsible for authenticating and
receives from the MS, but the RS does not process it. RS does
managing freshness of AK/TEK. The proposed scheme
not have any key information relevant to the MS, and all the
alleviates above security problems and examined how it
keys related to MS are maintained at the MS and MR-BS [13].
satisfies the security requirements of non-transparent Relay-
When the SA is established between RS and MR-BS in the
based WiMAX networks.
MR system, key data is shared and maintained at the particular
RS and MR-BS, such as AK, and the intermediate RS does not

32 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
VI. SEAKS NRS1 will be able to continuously exchange encrypted traffic
A. Authentication Procedures of NRS1 with MR-BS with the MR-BS.

Self organized and efficient authentication and key
management scheme (SEAKS) is based on self organized
model using non-transparent, decode and forward Relays.
SEAKS provides hybrid authentication scheme with
distributed authentication and localized re-authentication and
key maintenance. However, this technique not only helps in
minimizing the overall authentication overhead on MR-BS
and AAA server but also provides efficient way to
countermeasure the vulnerabilities; let’s consider any non-
transparent relay stations such as NRS1 wants to join the
WiMAX networks. NRS1 sends its Auth-REQ message to the
serving MR-BS, Auth-REQ includes manufacturer-issued
Figure 1: Authentication of NRS1 with MR-BS
X.509certificates, a description of cryptography algorithms
and NRS’s basic CID. The CID that assigned during the initial A TEK state machine remains active as long as NRS1 is
ranging, normally primary SAID is equal to the basic CID. In authorized to operate in the MR-BS security domain i.e. with
response to an authorization Request message, a MR-BS valid AK. NRS1 is authorized to participate in that particular
validates the requesting NRS’s identity, determines the security association [1] [2]. The parent authorization state
encryption algorithm and protocol support, activates an AK machine stops its entire child TEK state machines when NRS
for NRS1, encrypt it with the NRS1’s public key and send it receives from the MR-BS authorization reject during the
reauthorization cycle. We can say, this is localized
back to the NRS1 is AUTH-REP message. It also includes 4
authentication between NRS1 and MR-BS and these
bit sequence number, used to distinguish between successive procedures are same as mentioned in [3][4]. All the key state
generations of AKs, a life time, and the securities identities for machines are refreshing the keys. Now NRS1 is eligible to
which NRS1 are authorized to obtain keying materials. Once transmit UL-MAP message and any node listening to this
authenticated and obtain the authorization key (AK), NRS1 message, can sends the AUTH-REQ.
must periodically refresh its AK by reissuing an AUTH-REQ Now, there is another non-transparent relay station NRS2
message to the MR-BS. However, reauthorization is identical that wants to join the network. Due to its non-transparent
nature, it is not in the coverage of MR-BS and only NRS1 can
to authorization with the exception that NRS1 does not send
listen to it. According to SEAKS, NRS2 listened to the UL-
its authentication information messages during reauthorization MAP from NRS1 and sends the AUTH-REQ message to
cycle, to avoid service interruption during reauthorization, NRS1. However, any non-transparent node that wants to join
successive generations NRS1 AKs have overlapping lifetime. the network must have to authenticate itself with MR-BS as
Both NRS and MR-BS support up to two simultaneously MR-BS is directly attached to the AAA server, while NRS1
active AKs during these transition period. Authentication of cannot authenticate NRS2 on behalf of MR-BS.
NRS1 with MR-BS is shown in Figure 1.
Once NRS1 achieve authorization, its starts a separate B. Authentication Procedure of NRS2 with MR-BS
traffic encryption key (TEK) state machines for each of SAID
defined in the AUTH-REP message. Each TEK state machine According to SEAKS, NRS1 received the AUTH-REQ
operating within the NRS1 is responsible for managing the (NRS2) and send it to MR-BS during the refreshing of AK
keying material associated with its respective SAID. TEK message because these authentications are delay tolerance and
secured. NRS1 receive MACPDU of NRS2 and encapsulate it
state machine periodically send the key request messages to
into its own PKM-REQ message of type 9 and code 4 [1] [2].
the MR-BS to refresh the keying material for their respective MR-BS receives MACPDU of NRS1 which is basically sent
SAID. TEK is encrypted by appropriate KEK derived from the for refreshing AK. MR-BS will check MAC header of NRS1,
AK. The operation of the TEK state machine’s key request if RAR (Relay Auth Request) is equal to 1, it means there is
scheduling algorithm, combined with the MR-BS’s regimen one relay request inside MACPDU, RAR is basically the
for updating and using SAID keying materials ensure that reserve bit utilized for RAR indications.

33 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
Once MR-BS obtains AUTH-REQ of NRS2, it validates
its authenticity and activates AK2 and other parameters,
encrypt it with NRS1 public key and responds to NRS1 in its
AUTH-RSP message. NRS1 receives NRS2’s security info,
save one copy of all info into its knowledge shared table
(KST) generates AK21, encrypt it with NRS2 public key, and
sends its AUTH-RSP message to NRS2.

Figure 3: Authentication of NRSn with NRS1/MR-BS

means its PKM-AUTH–REQ message, once NRS2 receives
this message, it will check RAR values. If the value is one, it
will check inside the Mac payload and save the message to its
KST, then forward it to NRS1. Before sending, it will again
set the RAR==1. Hence, there are two Mac messages present
Figure 2: Authentication of NRS2 with NRS/MR-BS inside the Mac payload, one is AUTH-REQ (code 4) and the
other is KEY-REQ (code 5). NRS1 will receive this message
Once NRS2 get authenticated, it will start its separate and check RAR value; if it is one then it will copy the AUTH-
authorization and traffic encryption key state machine with REQ message to its KST, else it will ignore and forward it to
NRS1. As mentioned in the previous section, all the relays MR-BS. MR-BS will receive the message and validate it. MR-
involved are distributed, non-transparent, and decode and BS will send back the AUTH-RSP message with type 9. Again
forward. Thus, they can generate AUTH-RSP on behalf of here, there are two Mac messages inside the macpayload, one
MR-BS as shown in Figure 2. However, it cannot authenticate is with key reply (code 8) and other is auth-reply (code 5) to
its real validity because it does not have vendor’s digital NRS1. NRS1 check the code values, if it is 5, it will send to
certificate database. If NRS1 fails to re-authenticate before the NRS2. If 8 then it will use for its refreshing of keys. NRS2
expiration of its current AK, the MR-BS will hold no active again receives two Mac messages inside the payload, one is
AKs for NRS1 and will consider not only NRS1 but also all with code 5 and other is with code 8. It will retain code 8 with
others NRS unauthorized. A MR-BS will remove from its itself and send the code 5 message to NRS3. Thus NRS3 is
keying tables all TEKs associated with NRS1 [4] [12]. All authenticated with MR-BS with distributed manner and later it
NRSs maintain KST of recently exchanged AK with its will maintain its keys locally as mentioned in the previous
neighbours. If NRS2 fails to re-authenticate before the sections. The illustrations of authentication procedures of
expiration of its current AK, NRS1 will wait until it sends NRSn with MR-BS are shown in Figure 3.
AUTH-REQ message, NRS1 will check its KST, if it found
then it validates its authenticity locally rather than sending
D. Localized and Distributed Key Management in Relay-
again to MR-BS and wait for the response and compute the
Based IEEE 802.16 Network
keys and send to NRS2. The advantage is the communication
cost in shape of authentication overhead and thus less
complexity. We assume that all the NRS are authenticated and
maintains theirs KST. Inside the KST, we have two portions,
one is updated and other is non-updated stacks. All the active
C. Authentication Procedures of NRSn with NRS1/MR-BS and valid AK, TEK and SAIDLIST are residing inside the
updated one, and all the expired and revoked keys are inside
Now, if NRS3 wants to join the network, it will send the that non-updated stack. If any new NRS wants to join the
AUTH-REQ message to MR-BS, as it is working in non- network, the serving NRS first look at in its KST in updated
transparent mode. Hence, it has to send the request to the non- stack. If it cannot find the required information, it will move to
transparent and authenticated relay which should be inside its non-updated stack. If still it cannot find inside the non-updated
coverage that is NRS2. While sending the message, NRS3 will stack, the serving NRS will send the AUTH-REQ to the MR-
set RAR==1, inside the macheader so that NRS2 can BS through other NRS and all other procedures are the same.
recognize, there is one AUTH-REQ message inside the Mac The localized re-authentication and key maintenance
payload, and set the TYPE value ==8 and code ==4, which procedures is shown in Figure 4. If incase it found the

34 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
information in either of the stack, it validates its authenticity authentication protocol is illustrated in [3] [4]. We have
and send SAIDLIST and AK in AUTH-REP message and evaluated our scheme in terms of communication costs some
send one copy to the MR-BS for its own KST. key vulnerabilities and their countermeasures.
A. Communication Cost
The communication cost of our proposed scheme is mainly
comprises of re-authentication and key maintenance
overheads. The total communication costs of SEAKS can be
evaluated into two phases, the AUTH-REQ and AUTH-REP
phases. In AUTH-REQ phase, the source NRS sends its
AUTH-REQ as well as others NRS AUTH-REQs directly via
one hop to the MR-BS. This type of authentication occurs
once for specific NRS as after authentication, source NRS is
responsible for authenticating others NRSs who have already
obtained their AK/SAID. Within this first phase, we have
another issue of refreshing AK/TEK and all the NRS/MS have
Figure 4: Localized Re-Authentication and Key Maintenance to periodically and constantly send their refreshing request.
According to the standard, AK/TEK is refreshed by sending to
MR-BS validates its authenticity. If its valid then it will save the MR-BS with Multihop using Multihop Relays, but in our
in its KST else it will send AUTH-REJECT message in scheme, this is done localized as this system became
AUTH-REP. Now the entire network is doing distributed distributed. Hence, the communication cost of sending AUTH-
authentication as shown in Figure 5. REQ with refreshing AK/TEK can be calculated as follow

Figure 5 shows overall flow of our self organized re-
authentication and key management schemes in non- : 1 _
transparent Relay-based WiMAX network.
Where H is the average number of Hops between the source
and the destination, n is the number of NRS participating in
the entire network, certificate size is important parameter to be
counted as NRS also combine other AUTH-REQs with their
digital certificates.
In the AUTH-REP phase, MR-BS sends its AUTH-REP
message to its neighbor NRS with AK/SAID, this message is
unicast altogether with separate other AK/SAID for other
requesting NRS. Once NRS receives AK/SAID from MR-BS
it is encrypted with public key of requesting NRS, save the
copy to its local repository and send it back to requesting
NRS. The requesting NRS maintains it is AK/TEK with single
hop with serving NRS, thus minimize the authentication and
Figure 5: localized distribution of Keys using SEAKS key maintenance overhead, the communication cost of this
phase can be calculated as follows
Instead of re-authentication and refreshing keys with MR-BS
and gave birth to authentication and key maintenance
overhead, they create a very self-organized community to re-
authenticate and refresh keys to avoid delay and overheads. : 1 _
There is a very strong trust worthy and self-organized
environment is generated after the successful authentication of
all NRSs.
Hence, the total communication cost of AUTH-REQ and
AUTH-REP phases can be calculated as follows:
VII. ANALYSIS OF OUR PROPOSED APPROACH
In our proposed scheme, we used NRS’s manufacturer
certificates, capabilities, nonce and lists of SAID as sending 1 1 _
parameters and AK, life time of AK, its capabilities, nonce
and digital signatures as receiving parameters. The

35 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
B. Evaluation against Denial of Service & Reply Attack interleaving attacks to attempt, we assume that PKMv2
For the denial of service, this attack exists only on pre- protocols in used to authenticate the participating NRS and
authentication procedures. DoS and replay attacks are MS. Let’s say an adversary impersonates any NRSj and send
explained briefly in the previous section. The proposed the AUTH-REQ message to MR-BS, MR-BS will validates
scheme work well with Multihop non-transparent relay based and generate AK for adversary. But adversary cannot decrypt
WiMAX network. As there are numbers of NRS participating the AK because it do not have private key, it need to force
in environment thus it becomes fully self-organized after NRSj to send once again the AUTH-REQ. Previously, once
successful stability time. Let suppose, when an adversary NRSj send the AUTH-REQ, it set the time out value but
impersonate NRS and send AUTH-REQ message to MR-BS, within that value it have not received any authenticated
MR-BS validate its authenticity, generate AK, copy the message from MR-BS, it assume the link is broken or some
certificate in its KST and send the AUTH-RSP to an other technical error. NRSj will try to scan other UL-MAP,
adversary. Hence, adversary don’t have the private key of and will found let’s say NRSi, and will send the AUTH-REQ
NRS thus could not decrypt AK; it can only just reply this to MR-BS. MR-BS will reject legitimate NRSj request
message several time. Whenever, NRS send the AUTH-REQ because, and there is already certificate present in KST of
message to MR-BS, it usually set the time out value, and if the MR-BS. Again NRSj receive AUTH-REJECT message from
time out value reached to the limit, it sends the request again, NRSi, NRSj will set the time out value and again send the
here in this case, the time out value already reached to the AUTH-REQ via NRSi. There are two main reasons to adopt
limit, but there is no response from the MR-BS. NRS will the same path to authenticate itself, firstly, at least NRSj get
again search for UL-MAP, we assume that it will find another the response from this links, and secondly it assume to be due
path say NRSi, NRSi is inside the coverage of MR-BS, NRS to some technical errors. On the other hand, according to
will send the AUTH-REQ second time to NRSi, NRSi will SEAKS, after specific time out value, MR-BS have not get the
send the AUTH-REQ message to MR-BS, again MR-BS response from adversary, thus it will delete certificate of
validate the AUTH-REQ, generate the AK and send the NRSj. NRSj after time out, sends the AUTH-REQ again and
AUTH-RSP to NRSi and consequently NRS, NRS send will be authenticated and MR-BS will save its certificate in its
message III to MR-BS and thus get authenticated from the KST. By applying SEAKS and due to storage of AK/SAID in
MR-BS. Later NRS will start its AK and TEK refreshing with every NRS repositories, and NRS itself encrypt all the
NRSi. On the other hand, an adversary is still replaying the AK/SAID and TEK for others NRS, and due to distributed
message multiple times to exhaust the MR-BS. Now, MR-BS authenticated and localized re-authenticated and key
will again receive the AUTH-REQ message from adversary. maintenance, a very strong self-organized trustworthy
MR-BS knows that NRS is part of authenticated network and environment is created thus its quite impossible to get success
MR-BS is not expecting any message of AUTH-REQ from in interleaving attacks once the SEAKS got its stability.
this certificate. But if MR-BS receives any AUTH-REQ
message from the same certificates it will simply ignore this
VIII. CONCLUSION AND FUTURE WORK
message. After specific stability time, certificate of NRS is
shared with all the participating nodes, thus give maximum In this paper, we addressed a self organized efficient
protection against Do and Reply attacks. For adversary to authentication and key management scheme (SEAKS), hop-
transmit one way message several times without response need by-hop authentication and key management scheme in non-
some extra power, thus after some time adversary will stop transparent Relay-based WiMAX network. This scheme is
sending the message and the denial of service attempt became suitable for both fixed as well as mobile non-transparent
unsuccessful. As we mentioned previously, reply attack comes Relays. We have presented our security goals and stated
first and denial of service is the ultimate result of reply attack security analysis of proposed scheme to evaluate it against
where MR-BS after several reply attacks deny that particular those goals. SEAKS provides hybrid authentication scheme
certificate thus deny legitimate node. Hence, our scheme with distributed authentication and localized re-authentication
works well both denial of service and reply attack in a very and key maintenance. However, this technique not only helps
efficient manner. in minimizing the overall authentication overhead on MR-BS
and AAA server but also provides efficient way to
C. Evaluation against Interleaving Attack countermeasure the vulnerabilities In this scheme, NRS need
to first authenticate itself with MR-BS prior to accept AUTH-
To avoid Man-in-the-Middle attack, mutual authentication REQ from other NRS/MS once authenticated and get the
was provided and adds an additional message to provide NRS required AK/SAID, it continue its AK/TEK authorization state
acknowledgement and achieve X.509 three way machines to refresh above keys. After authenticated, it can
authentications, but this enhanced version is also vulnerable to start broadcasting UL-MAP to accept AUTH-REQ , after
an interleaving attack, which is explained in the previous receiving any AUTH-REQ it send it to MR-BS for validation,
section. The proposed scheme work well with Multihop non- MR-BS authenticate and send AK/SAID for particular request,
transparent relay based WiMAX network. As there are NRS receives and encrypt it with public key of requesting
numbers of NRS participating in environment thus it becomes NRS and send back. Now requesting NRS start authorization
fully self-organized after successful stability time. For state machines to refresh above keys with NRS, at any time,

36 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
all the NRS and MR-BS will maintain their local repositories. [13] D. Johnston and J. Walker, "Overview of IEEE 802.16 Security,"
IEEE Security and Privacy Magazine, vol. 2, no. 3, pp. 40-48,
If any NRS cannot refresh its key within particular given time May-June 2004..
due to uncertain circumstances, according to standard, it have [14] Adnan Shahid Khan , N.Fisal , N.N.M.I. Ma’arof , F.E.I.
to re-authenticate with MR-BS, but in our scheme, it will send Khalifa ,M. Abbas ,Security Issues and Modified Version of PKM
the request to NRS, NRS will look into its local repositories, if Protocol in Non-transparent Multihop Relay in IEEE 802.16j
found then send AK/SAID by itself it will send the AUTH- Networks, International Review on Computers and Software -
January 2011 (Vol. 6 N. 1 pp. 104-109).
REQ to MR-BS for authentication and validation and consider
[15] Xinmin Dai, Xiaoyao Xie, “Analysis and Research of Security
it as a new NRS/MS. Mechanism in IEEE 802.16j” Guizhou Normal University
In our future work, we will continue to implement a Guiyang, China, 2010
prototype of SEAKS and extend the scale of the experiments [16] Vamsi Krishna Gondi, “Security and Mobility architecture for
and to allow the emergence of other key management isolated wireless networks using Wimax as an Infrastructure”,
Network and Multimedia Systems Group, France, 2009
techniques to come up with highly efficient and secure key
management scheme in terms of throughput, complexicity,
and authentication overhead.
ACKNOWLEDGEMENT
The author would like to thanks to all WiMAX research group
ADNAN SHAHID KHAN received his degree of B.Sc
and especially sincerest gratitude to Ministry of Higher (Hons) in Computer Science from University of the
Education Malaysia under Malaysian Technical Cooperation Punjab, Lahore, Pakistan in 2005. Master of
Programme (MTCP) for their full support and Research Engineering degree in Electrical (Electronics &
Telecommunication) from Universiti Teknologi
Management Center (RMC), Universiti Teknologi Malaysia Malaysia, Skudai, Malaysia in 2008.Currently, he is
(UTM) and MIMOS BERHAD for their partial contribution. pursuing his PhD in Electrical Engineering at the
Faculty of Electrical Engineering, Universiti Teknologi
REFERENCES Malaysia, Skudai, 81310, Johor Bahru, under the supervision of Prof.
Dr. Norsheila Fisal .His current Research interests are in the area of
[1] IEEE Std 802.16-2009: Air Interface for Broadband Wireless Security Issues in IEEE 802.16 Protocol and Cognitive Radio
Access Systems, 2009 Networks. He is also student member of IEEE since 2007.
[2] IEEE Std 802.16j-2009, Amendment to IEEE STD 802.16-2009
[3] S. Xu and Huang. Attack on PKM protocols of IEEE 802.16 and its
later version. In international Symposium on wireless NORSHEILA FISAL received her B.Sc. in Electronic
Communication System (ISWCS), 2006. Communication from the University of Salford,
[4] Sen Xu, Manton Matthews and Chin-Tser Huang. Security Issues Manchester, U.K. in 1984. M.Sc. degree in
Telecommunication Technology, and PhD degree in
in Privacy and Key Management Protocols of IEEE 802.16. In
Data Communication from the University of Aston,
ACM SE'06. Florida USA. March 2006
Birmingham, U.K. in 1986 and 1993, respectively.
[5] Steven W.Peters and Robert W.Heath, Jr,”The Future of Wimax: Currently, she is the Professor with the Faculty of
Multihop Relaying with IEEE 802.16j”, IEEE communication Electrical Engineering, University Technology Malaysia and Director of
Magazine, January 2009. Telematic Research Group (TRG) Laboratory. Her current research interests
[6] Mosato Okuda, Chenxi Zhu and Dorin Viorel, Multihop Relay are in Wireless Sensor Networks, Wireless Mesh Networks, And Cognitive
Extension for Wimax Networks- Overview and Benefits of IEEE Radio Networks
802.16j Standard, FUJITSU Sci.Tech.J., 44,3, p.292-302 (July
2008)
[7] Adnan Shahid Khan et. al. “Efficient Distributed Authentication MAZLAN ABBAS received his B.Eng. in Electrical from
Key Scheme for Multi-hop Relay In IEEE 802.16j Network”, Universiti Teknologi Malaysia in 1984, M.Sc. In
International Journal of Engineering Science and Technology Telematics from Essex University in 1986, and PhD
(IJEST), Vol. 2(6), 2010, 2192-2199 degree in Telecommunications from Universiti Teknologi
[8] Taeshik Shon, Wook Choi: An Analysis of Mobile WiMAX Malaysia in 1992. Currently, he is the Chief Research
Security: Vulnerabilities and Solutions, First Director of Wireless Communications Cluster of MIMOS
InternationalConference, NBiS 2007, LNCS, Vol. 4650, pp. 88-97, Berhad and also the Adjunct Professor with the Faculty of
Electrical Engineering, Universiti Teknologi Malaysia.
2007.
His current research interests are in WiMAX, LTE, IMS and IPv6.
[9] Y.Lee, H.K.Lee, G.Y.Lee, H.J.Kim and C.K.Leong, “Design of
Hybrid Authentication Scheme and Key Distribution for Mobile
Multi-Hop Relay in IEEE 802.16j”, EATIS’09, June 3-5,
MAZLINA ESA received her BEE (Hons.), MSc in
Prague,CZ, 2009.
RF Engg., and PhD in Electrical and Electronics Engg.
[10] Huang C, Chang J. Responding to security issues in Wimax from Universiti Teknologi Malaysia, Univ. of Bradford
networks. IT Professional 2008; 10(5):15-21. (UK), and Univ. of Birmingham (UK), in 1984, 1987,
[11] Adnan Shahid Khan, Norsheila Fisal, Sharifah Kamilah, Rozeha A and 1996, respectively. She is currently a Professor
Rashid and M Abbas. Article: Secure and Efficient Multicast with the Faculty of Electrical Engg., UTM. Her
Rekeying Approach For Non-Transparent Relay-Based IEEE research interests include RF/microwave and antenna
802.16 Networks. International Journal of Computer engineering, THz/PHz technology, wireless power
Applications16(4):1–7, February 2011. Published by Foundation of transmission, cognitive radio, and qualitative research.
Computer Science She was the IEEE Malaysia AP/MTT/EMC Chapter Chair from 2007 to
Jan 2011, and currently the Counselor of IEEE UTM Student Branch.
[12] "Draft Standard for Local and Metropolitan Area Networks, She is an active Senior Member of IEEE.
Part16: Air Interface for Broadband Wireless Access Systems",
IEEE P802.16 Rev2/D9, January 2009

37 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
SHARIFAH KAMILAH BNT SYED YUSOF
received BSc (cum laude) in Electrical Engineering from
Geoge Washington University USA in 1988 and
obtained her MEE and Ph.D in 1994 and 2006
respectively from universiti Tecknologi Malaysia. She is
currently Associate Professor with the department of
Radio Communication, Faculty of Electrical Engineering Universiti
Teknologi Malaysia. Her research interest includes OFDMA based
system, Software define Radio and Cognitive radio.

SHARIFAH HAFIZAH SYED ARIFFIN Received her
B.Eng (Hons) from University North London in 1987, and
obtained her M.E.E and Ph.D in 2001 and 2006 from
Universiti Teknologi Malaysia, and Queen Marry
University. London respectively. She is currently Senior
lecturer with Faculty of Electrical Engineering, Universiti
Teknologi Malaysia. Her current research interest are in
Wireless sensor networks, IPV6, Handoff Management in Wimax,
6loWPAN and Network and Mobile Computing System.

38 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

A digital image encryption algorithm based on
chaotic logistic maps using a fuzzy controller
Mouad HAMRI #1 , Jilali Mikram #2 , Fouad Zinoun &3
#
Mathematics and computer science department, Science University of Rabat-Agdal
4 Avenue Ibn Battouta Rabat Morocco
&
Economical sciences and management department, University of Meknes Morocco
1
hamri.mouad@gmail.com
2
mikram@fsr.ac.ma
3
fouad.zinoun@gmail.com

Abstract—In this paper we will present a digital image encryp- and the quantum machines that can be a reality soon.
tion algorithm based on chaotic logistic maps and using fuzzy Chaotic dynamical systems present a very important tool to
logic (FL-CM-EA). Many papers was published in the recent build efﬁcient and secure cryptosystems thanks to their high
years about encryption algorithm using chaotic dynamical sys-
tems thanks to the set of very interesting properties guaranteed sensitivity to initial conditions, their ergodicity propriety, their
by these chaotic dynamical systems: high sensitivity to initial simplicity of implementation and also the very interesting
conditions, ergodicity, simplicity of implementation..., that can execution time that help to have a real-time applications.
be used to conceive efﬁcient cryptosystems. In this paper we propose an encryption algorithms using not
The main idea of this paper is the usage of a fuzzy logic set only one logistic map but a map of many logistic maps and
of rules to control the next iteration of our proposed iterative
mechanism using a set of logistic maps. the iterations are deﬁned using a set of fuzzy logic rules.
An introduction to chaotic dynamical systems and logistic map The rest of this paper will be as follow: section 2 introduces
is given followed by an introduction to fuzzy logic. A complete chaotic dynamical systems and logistic map, section 3
speciﬁcation of the proposed algorithm is presented with a set introduces fuzzy logic, section 4 presents the proposed
of security analysis tests that show the efﬁciency and the high algorithm with some results, section 5 presents the security
security level of the algorithm.
analysis tests and ﬁnally section 6 concludes this paper.
Keywords: cryptography, logistic map, fuzzy logic, image II. C HAOTIC DYNAMICAL SYSTEMS AND LOGISTIC MAP
encryption, security analysis, dynamical systems, chaos theory.
Roughly speaking, a dynamical system ([1-4],[11-12]) con-
I. I NTRODUCTION sists of two ingredients: a rule which is described by a set of
Today the community network applications in the internet equations and specify how the system evolves and an initial
are been used by billions of people around the world and this condition from which the system starts. It can be deﬁned
usage rate is growing continuously. This implies that more also as a system of equations describing the evolution of a
and more amounts of information is being transmitted over mathematical model where the model is fully determined by
the internet. The data being transmitted includes all kind of a set of variables.
information format: text, audio, video, image and a lot of The logistic map (that will be used in our algorithm) is a very
other special formats. famous discrete dynamical system used in many researches
Images are used widely in our daily life in almost our when dealing with dynamical systems and chaos. It is deﬁned
communications, these communications includes military on the set [0, 1] and can be written:
communications, banks transactions and many other
xn+1 = rxn (1 − xn )
communications where the security is really mandatory.
This lead to conclude that image security is a very important Where x0 represent the initial condition, n ∈ N and r is
topic in our internet communication world. positive real number.
In reality, there is no universal deﬁnition for chaotic
Many algorithm have been proposed in the last years dynamical systems. The following deﬁnition tries to deﬁne a
to solve these security issues, using the classical encryption chaotic dynamical system using three ingredients that almost
algorithms such as RSA or EL-Gamal or using the elliptic everyone would agree on.
curves. The problem with the previous algorithm is that their
security relies on the fact that it is not feasible with today’s Chaotic dynamical system: Let f : X → Y a function
machines to factorize a large number or to solve the discrete (X, Y ⊆ R).
logarithm problem but this may not be true in the near future ˙
The dynamical system x = f (x) is said to be chaotic if the
especially with the recent advances in machines performances following proprieties are satisﬁed:

39 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

1- Sensitive dependance on initial conditions:∀β > 0, For TS fuzzy rules and unlike Mandany fuzzy rules, TS fuzzy
∃ε > 0 there exists a point y0 ∈ X and k > 0, such that: rules deﬁne the output variables as a function of the input
| x0 − y0 |< β ⇒ | xk − yk |> ε. variables. If we take the same example as before, a TS fuzzy
2- Density of periodic orbits:The ensemble of periodic rule can be described as follow:
orbits: {x0 ∈ X, ∃k > 0, xk = x0 } is dense in X.
IF x1 in S1 and x2 in S1 THEN y1 = f (x1 , x2 ) and
3- Deterministic: means that the system has no random or
y2 = g(x1 , x2 )
noisy inputs or parameters.
Where f and g are two real functions of any type.
The deﬁnition above is applied to both discrete and In general, the steps followed to construct a fuzzy controller
continuous dynamical systems. are:
The logistic map is a chaotic dynamical system and presents 1) Identifying and naming the fuzzy inputs and outputs.
a very high sensitivity to initial conditions for r between 2) Creating the the fuzzy membership functions.
about 3.57 and 4 (approximatively). 3) Constructing the fuzzy rules (Mandany or TS rules).
Fig.1 shows the bifurcation diagram of the logistic map. 4) Deﬁning the defuzziﬁcation process (convert fuzzy out-
puts to crisp outputs).
The ﬁgure Fig.2 shows an example of a possible fuzzy
controller.

Fig. 1. Bifurcation diagram of the logistic map

III. F UZZY LOGIC

In the 1960s, Lotﬁ Zadeh invented fuzzy logic [16,17],
which combines the concepts of crisp logic and the
Lukasiewicz sets by deﬁning graded membership. One of
Zadehs main insights was that mathematics can be used to link
language and human intelligence. Many concepts are better
deﬁned by words than by mathematics, and fuzzy logic and its Fig. 2. Diagram of a fuzzy controller
expression in fuzzy sets provide a discipline that can construct
better models of reality.
In the next section, we will present our encryption algorithm
Fuzzy logic is a form of many-valued logic in the opposite of
and we will describe all the parameters of the used fuzzy
the crisp logic which is a two-valued logic (binary logic).
controller.
Fuzzy logic involves linguistic variables with a truth value
in the interval [0, 1], it involves also fuzzy sets and fuzzy IV. T HE ALGORITHM
inference.
Every fuzzy model uses fuzzy rules which are linguistic if- The proposed algorithm (FL-CM-EA) takes as inputs a
then statements. These rules are linking the inputs variables plain-image P and a 128 bits key K then generates as output
to the output variables, they simply deﬁne the control logic. the cipher-image C.
Two major types of fuzzy rules exist: Mandany fuzzy rules and The main idea of the algorithm was to use not only a simple
Takagi-Sugeno (TS) fuzzy rules. An example of a Mandany logistic map to generate the encryption (decryption key) but to
fuzzy rule for a fuzzy system with two inputs and two outputs use what we have called ”fuzzy-logistic-map”, which is also
can be described as follow: a function from the interval [0, 1] to itself, using three fuzzy
rules and three logistic map (we can use as many logistic maps
IF x1 in S1 and x2 in S1 THEN y1 in S3 and y2 in S4 and fuzzy rules as we want but in this paper we will use three).

40 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

IT ER
If we call the three logistic maps LM1 , LM2 and LM3 then – KFi =( l=1 F M L(i + l)2 ) ×256 mod 256.
the fuzzy rules are as follow: – Run the FLM generator and stop after IT ER.
1) IF x IS M1 THEN FLM(x)=LM1 (x) = r1 x(1 − x)
2) IF x IS M2 THEN FLM(x)=LM2 (x) = r2 x(1 − x) • Step 3:Using the generated key, we will generate the
3) IF x IS M3 THEN FLM(x)=LM3 (x) = r3 x(1 − x) image C as follow:
For the rest of this paper, we will use the following values: – C0 (R) = (P0 (R) + KF0 ) mod 256.
r1 = 3.95, r2 = 3.9 and r3 = 3.8. – C0 (G) = (P0 (G) + KF0 ) mod 256.
The fuzzy sets M1 , M2 and M3 membership functions f1 , – C0 (B) = (P0 (B) + KF0 ) mod 256.
f2 and f3 are deﬁned as follow: and:
1
For i in [2, n]:
−2x + 1 if 0≤x≤ 2
f1 (x) = 1 – Ci (R) = (Pi (R) + KFi + Ci−1 (R)) mod 256.
0 if 2 ≤x≤1 – Ci (G) = (Pi,j (G) + KFi,j + Ci−1 (G))) mod 256.
2x if 0≤x≤ 1
2
– Ci (B) = (Pi,j (B) + KFi,j + Ci−1 (B))) mod 256.
f2 (x) = 1
−2x + 2 if 2 ≤x≤1 • Step 4: We reverse the data of the image C :
For i in [1, n]:
0 if 0≤x≤ 1
2
f3 (x) = 1 – Ci = Cn−i+1
2x − 1 if 2 ≤x≤1
• Step 5: ﬁnally we construct the cipher-image C by
For the defuzziﬁcation process, we use a center average repeating the step 3 using the image C :
defuzziﬁer and the crisp value of FLM(x) is:
– C0 (R) = (C0 (R) + KF0 ) mod 256.
3
i=1 µi LMi (x) – C0 (G) = (C0 (G) + KF0 ) mod 256.
F M L(x) = 3 – C0 (B) = (C0 (B) + KF0 ) mod 256.
i=1 µi
and:
Where µi represents the degree of membership of x in Mi . For i in [2, n]:
Before presenting the algorithm, the following notations are
– Ci (R) = (Ci (R) + KFi + Ci−1 (R)) mod 256.
presented:
– Ci (G) = (Ci,j (G) + KFi,j + Ci−1 (G))) mod 256.
P plain-image – Ci (B) = (Ci,j (B) + KFi,j + Ci−1 (B))) mod 256.
K 128 bits key • End
C cipher-image
Pi ith pixel of P
The decryption algorithm is identical to the encryption algo-
Pi (R, GorB) Red, Green or Blue value of the pixel i
F LMi fuzzy-logistic-map value after i iteration rithm, it receives as inputs the cipher-image C and the 128
Li (x0 , N ) Value of the logistic map i bits key K (the same used for the encryption) and returns as
starting from x0 after N iterations output the plain-image P.
F A map from the set of 32 bytes The only difference between the two algorithm is the step
numbers to the interval [0, 1] 3 and step 5 which are deﬁned as below for the decryption
algorithm.
The encryption algorithm description can be summarized • Step 3:
as following:
– C0 (R) = (C0 (R) − KF0 ) mod 256.
– C0 (G) = (C0 (G) − KF0 ) mod 256.
• Begin: – C0 (B) = (C0 (B) − KF0 ) mod 256.
• Step 1: We begin by generating an initial condition and:
x0 ∈ [0, 1]: For i in [2, n]:
x0 = F (K).
– Ci (R) = (Ci (R) − KFi − Ci−1 (R)) mod 256.
– Ci (G) = (Ci (G) − KFi − Ci−1 (G)) mod 256.
• Step 2: In this step we generate a key vector KF of
– Ci (B) = (Ci (B) − KFi − Ci−1 (B)) mod 256.
size n where n is the number of pixels of P using the
function getKey: • Step 5:

KF = getKey(x0 ). – P0 (R) = (C0 (R) − KF0 ) mod 256.
The function getKey is deﬁned as bellow: – P0 (G) = (C0 (G) − KF0 ) mod 256.
– P0 (B) = (C0 (B) − KF0 ) mod 256.
Run the FLM generator and stop after IT ER and:
iterations (the initial value is x0 and IT ER is an For i in [2, n]:
iteration parameter). – Pi (R) = (Ci (R) − KFi − Ci−1 (R)) mod 256.
For i in [1, n]: – Pi (G) = (Ci (G) − KFi − Ci−1 (G)) mod 256.
– Pi (B) = (Ci (B) − KFi − Ci−1 (B)) mod 256.

41 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

In the next section, we will present the security analysis going to be this time :(C1 (i, j),C2 (i, j)).
tests performed on our algorithm. Other measures are going to be used to compare two images
C1 and C2 as the Number of Pixels Change Rate (NPCR)
V. S ECURITY ANALYSIS
deﬁned as below:
In this section we will discuss the security analysis of
i,j D(i, j)
our algorithm such as key space analysis, sensitivity analysis N CP R = × 100%
n
(with respect to both the key and the plain-image) and ﬁnally
Where n is the images size (number of pixels) and:D(i, j) = 0
statistical analysis as any robust encryption algorithm should
if C1 (i, j) = C2 (i, j) and D(i, j) = 1 otherwise.
resist these attacks.
The Uniﬁed Average Changing Intensity (UACI) will be used
The computation was done using a PC with the following
as well and it is deﬁned as:
characteristics: 1,8GHz Core(TM) 2 Duo, 1.00 Go RAM and
120 Go hard-disk capacity. 1 C1 (i, j) − C2 (i, j)
U ACI = × 100%
n i,j 255
A. Key space analysis
The used key for our algorithm is a 128 bits key which Here C1 (i, j) and C1 (i, j) are grey-scale values of the images
means that we have 2128 possibilities to generate a secret key. pixels.
With such large key space, the encryption algorithm can be
1) Key sensitivity analysis: Key sensitivity is a required
considered secured. In addition to that, the chaotic system that
property to ensure the security of any image encryption
we are using to generate the cipher-image is highly sensitive
algorithm against some brute-force attacks.
to initial condition which will guarantee that having this large
To test the key sensitivity of the proposed algorithm,
key space both key and plain-image attacks will not affect the
we have generated randomly an encryption key:
security of the algorithm as we will see in the next sections.
”0CDA03C2D734F06C48A33ECBE3178632” then we
B. Sensitivity analysis encrypted an original image P using this key to obtain the
An efﬁcient image encryption algorithm should be highly image C1.
sensitive to the secret key and to the plaint-image, which We then slightly modiﬁed the key by
means that a single bit change in the encryption key will lead changing the most signiﬁcant bit to obtain:
to a very different cipher-image from the initial cipher-image ”8CDA03C2D734F06C48A33ECBE3178632”, and using
and similarly, only a pixel change in the plaint-image should this key we’ve encrypted the same original P message the
lead to a very different cipher-image from the initial cipher- obtain image C2.
image. Finally, we did the same as the last operation but
We will present in this section the results obtained by changing changing the least signiﬁcant bit to obtain the key:
one bit in the encryption key and one pixel in the plain-image ”0CDA03C2D734F06C48A33ECBE3178633” and using
and we will see the effects on the cipher-image. this last key we encrypted the original image P to obtain the
Before starting our analysis, we will introduce some famous image C3 (see ﬁgure Fig.3).
statistical measures that we will use in the next sections.
The ﬁrst measure that we will talk about is the statistical corre-
lation between two vertically adjacent pixels, two horizontally
adjacent pixels and two diagonally adjacent pixels.
To compute this measure, we ﬁrst take randomly a set of
adjacent pixels (vertically, horizontally or diagonally) from the
image (let’s say 1000 pairs) then we calculate their correlation
using the formulas: Fig. 3. From the left to the right: original image P, C1, C2 and C3
cov(x, y)
rxy =
D(x) D(y) We have calculated the correlation, the NCPR and the
UACI of each two of the three cipher-images C1, C2 and
Where:
N C3 (Table I, II and III).
1
E[x] = xi
N i=1 For the obtained results we can see clearly that a negligible
N correlation exists among the three images even if they was
1
D[x] = (xi − E[x])2 produced using the same original image and with a slightly
N i=1 different keys. We can see also that the rate of change NPCR,
the intensity of change UACI are really high,then we can
cov(x, y) = E[(x − E[x])(y − E[y])]
conclude that our algorithm is very sensitive to encryption
We will use also to compare two images C1 and C2 , their key change.
correlation deﬁned as above but the used pairs of pixels are

42 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Image 1 Image 2 Correlation
C1 C2 -0.000291
C1 C3 0.000004
C2 C3 -0.001109
TABLE I
C ORRELATION BETWEEN THE IMAGES C1, C2AND C3 OBTAINED BY
SLIGHTLY CHANGING THE ENCRYPTION KEY ( ONE BIT CHANGE )

Fig. 4. The image P1 (left) and the image C1 (right)
Image 1 Image 2 NCPR
C1 C2 99.6037%
C1 C3 99.6077%
C2 C3 99.6039%
TABLE II
NCRP OF THE IMAGES C1, C2 AND C3OBTAINED BY SLIGHTLY
CHANGING THE ENCRYPTION KEY ( ONE BIT CHANGE )

Image 1 Image 2 UACI
C1 C2 49.8139% Fig. 5. The image P2 (left) and the image C2 (right)
C1 C3 49.7397%
C2 C3 49.8706%
TABLE III
UACI OF THE IMAGES C1, C2 AND C3 OBTAINED BY SLIGHTLY
CHANGING THE ENCRYPTION KEY ( ONE BIT CHANGE )

2) Plain-image sensitivity analysis: After studying the key
sensitivity of the proposed image encryption algorithm, we Fig. 6. The image P3 (left) and the image C3 (right)
will study now its plaint-image sensitivity.
The algorithm should be also sensitive to any small change in Image 1 Image 2 Correlation
the plaint-image which means that changing only one pixel in C1 C2 -0.0840
C1 C3 -0.0192
the plaint-image should lead to a very different cipher-image. C2 C3 -0.0377
This property will guarantee the security of the algorithm
TABLE IV
against plaint-image brute-force attacks. C ORRELATION BETWEEN THE IMAGES C1, C2 AND C3 OBTAINED BY
To test the sensitivity to plaint-image, we will take an CHANGING ONLY ONE PIXEL OF THE ORIGINAL IMAGE
original image (P1) then we will encrypted it (we call
the cipher-image C1), and we will randomly change a
pixel in the original message then will encrypt the im- Image 1 Image 2 NCPR
C1 C2 99.6825%
age again (P2) to obtain a new cipher-image C2. We re- C1 C3 99.8608%
peat this a last time again to obtain a new image (P3) C2 C3 99.8608%
and a third cipher-image C3 (we have used as encryp- TABLE V
tion key:”0CDA03C2D734F06C48A33ECBE3178632”) (see NCRP OF THE IMAGES C1, C2 AND C3 OBTAINED BY BY CHANGING
ONLY ONE PIXEL OF THE ORIGINAL IMAGE
Fig.4, Fig.5 and Fig.6)
As we did for the previous section, we will calculate the
correlation, the NPCR and the UACI between each two of Image 1 Image 2 UACI
the three cipher-images (Tables IV, V and VI). C1 C2 51.3027%
Again, the obtained results show that a negligible correlation C1 C3 53.2280%
C2 C3 46.2083%
exists between the three cipher-images and we can see also that
the rate of change (NPCR) and the intensity of change UACI TABLE VI
UACI OF THE IMAGES C1, C2 AND C3 OBTAINED BY CHANGING ONLY
are really high. Form the previous results, we can conclude that ONE PIXEL OF THE ORIGINAL IMAGE
our algorithm is very sensitive also to plain-image change.

C. Statistical analysis
After studying the security of the proposed algorithm image P that will be encrypted to obtain a
against some brute-force attacks (key sensitivity and plain- cipher-image C (we have used also as encryption
image sensitivity), we will study in this section the security key:”0CDA03C2D734F06C48A33ECBE3178632”). We
against statistical attacks. then compare their histograms and compute for each image
To perform this study, we will consider an original the values of its two vertically adjacent pixels correlation, two

43 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

horizontally adjacent pixels correlation and two diagonally VI. C ONCLUSION
adjacent pixels correlation. In this paper we presented a digital image encryption
algorithm based on chaotic logistic maps and using a fuzzy
controller (FL-CM-EA).
The introduction of the fuzzy controller helped to use a set
1) Histogram comparisons: Fig.7 and Fig.8, presents the of logistic maps instead of one logistic map and therefore
histograms of the images P and C. increased the randomness of the generated inputs.
We have tested also the robustness and efﬁciency of the
proposed algorithm by performing a set of security analysis as
the key space analysis, the key sensitivity and the plaint-image
sensitivity analysis and some other statistical analysis as the
histogram and the pixels adjacent correlation analysis and
all the results demonstrated the high level of security of the
proposed algorithm.

R EFERENCES
Fig. 7. The image P and its histogram [1] V.I. Arnold, Ordinary Differential Equations (Springer, Berlin, 1977,
1992).
[2] Huaguang Zhang, Derong Liu, Zhiliang Wang, Controlling Chaos: Sup-
pression, Synchronization and Chaotiﬁcation (Springer, London, 2009).
[3] James D. Meiss, Differential Dynamical Systems (SIAM 2007).
[4] V.I. Arnold, Geometrical Methods in the Theory of Differential Equa-
tions (Springer, Berlin, 1988).
[5] V.I. Arnold and D.V. Anosov, Dynamical Systems I (Springer, Berlin,
1988).
[6] Yu. A. Mitropolsky and A.K. Lopatin, Nonlinear Mechanics, Groups
and Symmetry (Kluwer, Dordrecht, 1995).
[7] Yu. N. Bibikov, Local Theory of Analytic Ordinary Differential Equa-
tions, Lecture Notes in Mathematics (Springer, Berlin, 1979).
[8] A.D. Bruno, Local Methods in Nonlinear Differential Equations
Fig. 8. The image C and its histogram (Springer, Berlin, 1989).
[9] V.I Arnold, Mathematical methods of classical mechanics, (2nd edition,
Springer, 1989).
[10] Bellman R., Stability theory of differential equations , (MGH, 1953).
We can see clearly that the histogram of the image C is [11] R.C. Schmitt and K. Thompson, Nonlinear Analysis and Differential
Equations, An Introduction (Aug 2000).
almost uniform and very different from the histogram of the [12] P. Hartman, Ordinary Differential Equations (John Wiley and Sons
image P. This result conﬁrms that statistical attacks based 1964)
on the histogram analysis can’t give any clue to break the [13] George D.Birkhoff, Dynamical systems (American Mathematical Soci-
ety 1991)
algorithm as all the statistical information of the image P are [14] Floriane Anstett, Les systemes dynamiques chaotiques pour le chiffre-
lost after the encryption. ment : synthese et cryptanalyse (These) (Universite Henri Poincare -
Nancy 1)
[15] A. Menezes, P. van Oorschot, and S. Vanstone Handbook of Applied
Cryptography (CRC Press, 1996)
2) Adjacent pixels correlation comparisons: The last [16] F. Martin McNeill, Ellen Thro, Fuzzy logic a practical approach (AP
PROFESSIONAL, 1994)
statistical analysis performed is the adjacent pixels correlation. [17] Zadeh, L, Fuzzy sets, Information and Control (1965)
We calculate for each image (P and C) the three adjacent [18] Kamyar Mehran, Takagi-Sugeno Fuzzy Modeling for Process Control
pixels correlations: vertically, horizontally and diagonally. (Newcastle University 2008)
[19] Hossam El-din H. Ahmed, Hamdy M. Kalash, and Ossam S. Farag
The table VII shows the obtained results. Allah, An efﬁcient Chaos-Based Feedback Stream cipher (ECBFSC) for
Image Cryptosystems (SITIS 2006)
[20] Mouad HAMRI, Jilali Mikram and Fouad Zinoun. Chaotic Hash Func-
Image H Adj Corr V Adj Corr D Adj Corr tion Based on MD5 and SHA-1 Hash Algorithms (IJCSIS Vol. 8 No. 9
DEC 2010)
P 0.7773 0.8895 0.7507
C -0.0055 0.0093 -0.0007
TABLE VII
VERTICALLY, HORIZONTALLY AND DIAGONALLY ADJACENT PIXELS
CORRELATION OF THE IMAGES P AND C

From the obtained results, we can see clearly that the pixels
of the plait-image P are strongly correlated while a negligible
correlation exists between those of the cipher-image C.
This result shows again that the proposed algorithm can be
considered as secure against statistical attacks.

44 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Volume 9 No. 3, March 2011

Performance Analysis of Connection Admission
Control Scheme in IEEE 802.16 OFDMA Networks
Abdelali EL BOUCHTI, Said EL KAFHALI and Abdelkrim HAQIQ

Computer, Networks, Mobility and Modeling laboratory
e- NGN research group, Africa and Middle East
FST, Hassan 1st University, Settat, Morocco
Emails: {a.elbouchti, kafhalisaid, ahaqiq} @gmail.com

Abstract—IEEE 802.16 OFDMA (Orthogonal Frequency Division and also it is robust to inter-symbol interference and
Multiple Access) technology has emerged as a promising frequency-selective fading. OFDMA has been adopted as the
technology for broadband access in a Wireless Metropolitan Area physical layer transmission technology for IEEE
Network (WMAN) environment. In this paper, we address the 802.16/WiMAX-based broadband wireless networks.
problem of queueing theoretic performance modeling and
Although the IEEE 802.16/WiMAX standard [12] defines the
analysis of OFDMA under broad-band wireless networks. We
consider a single-cell IEEE 802.16 environment in which the base physical layer specifications and the Medium Access Control
station allocates subchannels to the subscriber stations in its (MAC) signaling mechanisms, the radio resource management
coverage area. The subchannels allocated to a subscriber station methods such as those for Connection Admission Control
are shared by multiple connections at that subscriber station. To (CAC) and dynamic bandwidth adaptation are left open.
ensure the Quality of Service (QoS) performances, a Connection However, to guarantee QoS performances (e.g., call blocking
Admission Control (CAC) scheme is considered at a subscriber rate, packet loss, and delay), efficient admission control is
station. A queueing analytical framework for these admission necessary in a WiMAX network at both the subscriber and the
control schemes is presented considering OFDMA-based base stations.
transmission at the physical layer. Then, based on the queueing
The admission control problem was studied extensively for
model, both the connection-level and the packet-level
performances are studied and compared with their analogues in wired networks (e.g., for ATM networks) and also for
the case without CAC. The connection arrival is modeled by a traditional cellular wireless systems. The classical approach
Poisson process and the packet arrival for a connection by a two- for CAC in a mobile wireless network is to use the guard
state Markov Modulated Poisson Process (MMPP). We channel scheme [5] in which a portion of wireless
determine analytically and numerically different performance resources (e.g., channel bandwidth) is reserved for handoff
parameters, such as connection blocking probability, average traffic. A more general CAC scheme, namely, the fractional
number of ongoing connections, average queue length, packet guard scheme, was proposed [13] in which a handoff
dropping probability, queue throughput and average packet call/connection is accepted with a certain probability. To
delay.
analyze various connection admission control algorithms,
Keywords-component: WiMAX, OFDMA, MMPP, Queueing analytical models based on continuous-time Markov chain,
Theory, Performance Parameters. were proposed [4]. However, most of these models dealt only
with call/connection-level performances (e.g., new call
I. INTRODUCTION blocking and handoff call dropping probabilities) for the
traditional voice-oriented cellular networks. In addition to the
The evolution of the IEEE 802.16 standard [14] has spurred connection-level performances, packet-level (i.e., in-
tremendous interest from the network operators seeking to connection) performances also need to be considered for data-
deploy high performance, cost-effective broadband wireless oriented packet-switched wireless networks such as WiMAX
networks. With the aid of the Worldwide Interoperability for networks.
Microwave Access (WiMAX) organization [1], several An earlier relevant work was reported by the authors in
commercial implementations of WiMAX cellular networks [10]. They considered a similar model in OFDMA based-
have been launched, based on OFDMA for non-line-of-sight IEEE 802.16 but they modeled both the connection-level and
applications. The IEEE 802.16/WiMAX [2] can offer a high packet-level by tow different Poisson processes and they
data rate, low latency, advanced security, quality of service compared various QoS measures of CAC schemes. In [15], the
(QoS), and low-cost deployment. authors proposed a Discrete-Time Markov Chain (DTMC)
OFDMA is a promising wireless access technology for the framework based on a Markov Modulated Poisson Process
next generation broad-band packet networks. With OFDMA, (MMPP) traffic model to analyze VoIP performance. The
which is based on orthogonal frequency division multiplexing MMPP processes are very suitable for formulating the multi-
(OFDM), the wireless access performance can be substantially user VoIP traffic and capturing the interframe dependency
improved by transmitting data via multiple parallel channels, between consecutive frames.

45 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Volume 9 No. 3, March 2011
In this paper, we present a connection admission control scheme for subscriber stations are proposed. A threshold C is
scheme for a multi-channel and multi-user OFDMA network, used to limit the number of ongoing connections. When a new
in which the concept of guard channel is used to limit the connection arrives, the CAC module checks whether the total
number of admitted connections to a certain threshold. A number of connections including the incoming one is less than
queueing analytical model is developed based on a three- or equal to the threshold C. If it is true, then the new
DTMC which captures the system dynamics in terms of the connection is accepted, otherwise it is rejected.
number of connections and queue status. We assume that the
connection arrival and the packet arrival for a connection III. FORMULATION OF THE ANALYTICAL MODEL
follow a Poisson process and a two-state MMPP process
respectively. Based on this model, various performance A. Formulation of the Queueing Model
parameters such as connection blocking probability, average An analytical model based on DTMC is presented to
number of ongoing connections, average queue length, analyze the system performances at both the connection-level
probability of packet dropping due to lack of buffer space, and at the packet-level for the connection admission schemes
queue throughput, and average queueing delay are obtained. described before. We assume that packet arrival for a
The numerical results reveal the comparative performance connection follows a two-state MMPP process [3] which is
characteristics of the CAC and the without CAC algorithms in identical for all connections in the same queue. The connection
an OFDMA-based WiMAX network. inter-arrival time and the duration of a connection are assumed
The remainder of this paper is organized as follows. to be exponentially distributed with average 1/  and 1/  ,
Section II describes the system model including the objective respectively.
of CAC policy. The formulation of the analytical model for
An MMPP is a stochastic process in which the intensity of
connection admission control is presented in Section III. In a Poisson process is defined by the states of a Markov chain.
section IV we determine analytically different performance That is, the Poisson process can be modulated by a Markov
parameters. Numerical results are stated in Section V. Finally, chain. As mentioned before, an MMPP process can be used to
section VI concludes the paper. model time-varying arrival rates and can capture the inter-
frame dependency between consecutive frames ([6], [7], [8]).
The transition rate matrix and the Poisson arrival rate matrix of
II. MODEL DESCRIPTION
the two-state MMPP process can be expressed as follows:
A. System model  q q01   0 0
QMMPP   01  , =  (1)
We consider a single cell in a WiMAX network with a base
 q10 q10  0 1 

station and multiple subscriber stations (Figure 1). Each
subscriber station serves multiple connections. Admission The steady-state probabilities of the underlying Markov chain
control is used at each subscriber station to limit the number of are given by:
ongoing connections through that subscriber station. At each q10 q01
subscriber station, traffic from all uplink connections are ( MMPP ,0 ,  MMPP ,1 )  ( , ) (2)
aggregated into a single queue [11]. The size of this queue is q01  q10 q01  q10
finite (i.e., L packets) in which some packets will be dropped if The mean steady state arrival rate generated by the MMPP is:
the queue is full upon their arrivals. The OFDMA transmitter at q  q 
MMPP   MMPP  T  10 0 01 1 (3)
the subscriber station retrieves the head of line packet(s) and q01  q10
transmits them to the base station. The base station may
where  is the transpose of the row vector   (0 , 1 ) .
T
allocate different number of subchannels to different subscriber
stations. For example, a subscriber station with higher priority The state of the system is described by the
could be allocated more number of subchannels.
process X t  ( X , X t1 , X t2 ) , where X is the state of an
irreducible continuous time Markov chain and X t1
2
(respectively X t ) is the number of packets in the aggregated
queue (the number of ongoing connections) at the end of every
time slot t.
Figure 1. System model Thus, the state space of the system is given by:
E  {(i, j , k ) / i  {0,1}, 0  j  L, k  0} .
B. CAC Plicy .
The main objective of a CAC mechanism is to limit the For the CAC algorithm, the number of packet arrivals
number of ongoing connections/flows so that the QoS depends on the number of connections. The state transition
diagram is shown in (Figure 2). Here, (0 , 1 ) and  denote
performances can be guaranteed for all the ongoing
connections. Then, the admission control decision is made to
accept or reject an incoming connection. To ensure the QoS rates and not probabilities.
performances of the ongoing connections, the following CAC

46 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Volume 9 No. 3, March 2011
Note that the probability that n Poisson events with average C. Transition Matrix for the Queue
rate  occur during an interval T can be obtained as follows: The transition matrix P of the entire system can be
expressed as follows. The rows of matrix P represent the
e T ( T )n number of packets (j) in the queue.
fn ( )  (4)
n!
 p 0,0  p0 , A 
This function is required to determine the probability of  
both connection and packet arrivals.      
 p R ,0  pR ,R  pR ,R  A  (7)
P 
      
 p j, j R  p j, j  p j, jR 
 

       
Matrices p j , j ' represent the changes in the number of
packets in the queue (i.e., the number of packets in the queue
changing from j in the current frame to j ' in the next frame).
We first establish matrices v (i , j ),(i , j ') , where the diagonal
elements of these matrices are given as follows.
For r  {0,1, 2,..., D} and n  {0,1, 2,..., (k  A)}, l  1, 2,..., D ,
and m  1,2,...,(k  A) . The non-diagonal elements of
Figure 2. State transition diagram of discrete time Markov
chain. v (i , j ),(i , j ') are all zero.

B. CAC Algorithm  v (i , j );(i , j l ) 
  k 1,k 1  
n  r l
f n ( k i )[ R]r
In this case, the transition matrix Q for the number of
connections in the system can be expressed as follows:  v (i , j );(i , j  m ) 
  k 1, k 1  
r n m
f n ( k i )[ R]r (8)

 q 0 ,0 q 0 ,1 
 k 1,k 1   f n ( k i )[ R]r
   v (i , j );(i , j ) 
 q 0 ,1 q 1,1 q 1,2  
r n
Q      (5)
  Here A is the maximum number of packets that can arrive
 q C  2,C 1 q C 1,C 1 q C 1,C  from one connection in one frame, R indicates the maximum
  number of packets that can be transmitted in one frame
 q C 1,C q C ,C 
and D is the maximum number of packets that can be
where each row indicates the number of ongoing connections. transmitted in one frame by all of the allocated subchannels
As the length of a frame T is very small compared with allocated to that particular queue and it can be obtained from
connection arrival and departure rates, we assume that the D  min (R, j) . This is due to the fact that the maximum
maximum number of arriving and departing connections in a
number of transmitted packets depends on the number of
frame is one. Therefore, the elements of this matrix can be
obtained as follows: packets in the queue and the maximum possible number of
transmissions in one frame. Note that,  v(i, j );(i, j l )  ,
 k 1,k 1
qk,k1  f1() (1 f1(k)), k=0,1,...,C-1
v(i, j );(i, jm) 
 k1,k 1
and v(i, j);(i, j ) 
 k 1,k 1
represent the probability that
qk,k1  (1 f1())  f1(k), k=1,2,...,C (6)
the number of packets in the queue increases by n, decreases
qk,k  f1()  f1(k)  (1 f1()) (1 f1(k)), k=0,1,...,C by m, and does not change, respectively, when there are k
ongoing connections. Here,  v  denotes the element at row i
i, j
where qk ,k 1 , qk ,k 1 and qk ,k represent the cases that the and column j of matrix v, and these elements are obtained
number of ongoing connections increases by one, decreases by based on the assumption that the packet arrivals for the
one, and does not change, respectively. ongoing connections are independent of each other.
Finally, we obtain the matrices p j , j ' by combining both the
connection-level and the queue-level transitions as follows:

p j , j '  Qv ( i , j ),(i , j ') (9)

47 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Volume 9 No. 3, March 2011
IV. PERFORMANCE PARAMETERS
1 C L A
C 
In this section, we determine the connection-level and the
Ndrop   j1  [ p j , j m ]k ,l .(m  (L  j)). (i, j, k ) (14)
packet-level performance parameters (i.e., connection blocking i 0 k 1 j  0 m L   l 1 
probability, average number of ongoing connections in the
where the term  [ p j, jm ]k ,l  indicates the total probability
system, and average queue length) for the CAC scheme. C

 
These performance parameters can be derived from the  l 1 
steady state probability vector of the system states  , which is that the number of packets in the queue increases by m at
obtained by solving  P   and  1  1 , where 1 is a column every arrival phase. Note that, we consider probability
matrix of ones. pj, jm rather than the probability of packet arrival as we have to
Also, the size of the matrix P needs to be truncated at L consider the packet transmission in the same frame as well.
(i.e., the maximum number of packets in the queue) for the
After calculating the average number of dropped packets
scheme.
per frame, we can obtain the probability that an incoming
The steady-state probability, denoted by  (i, j , k ) for the packet is dropped as follows:
state that there are k connections and j  {0,1,..., L} packets N drop
in the queue, can be extracted from matrix  as follows: pdrop  (15)

 (i, j, k )   i j((C 1) k ) , i  0,1; k  0,1,..., C (10)
where  is the average number of packet arrivals per frame
and it can be obtained from
A. Connection Blocking Probability
This performance parameter indicates that an arriving   MMPP N k . (16)
connection will be blocked due to the admission control
decision. It indicates the accessibility of the wireless service E. Queue throughput
and can be obtained as follows: It measures the number of packets transmitted in one frame
1 L and can be obtained from
pblock    (i, j , C ). (11)   MMPP (1  pdrop ). (17)
i  0 j 0

The above probability refers to the probability that the F. Average Packet Delay
system serves the maximum allowable number of ongoing
It is defined as the number of frames that a packet waits in
connections.
the queue since its arrival before it is transmitted. We use
Little’s law [9] to obtain average delay as follows:
B. Average Number of Ongoing Connections
It can be obtained as Nj
D (18)
1 L C

N k   k . (i, j , k ) (12)
i0 j 0 k 0

V. NUMERICAL RESULTS
C. Average Queue Length Average In this section we present the numerical results of CAC
It is given by scheme. We use the Matlab software to solve numerically and
1 C L
to evaluate the various performance parameters.
N j   j. (i, j , k ) (13)
i 0 k 0 j 0
A. Parameter Setting
As in [10], we consider one queue (which corresponds to a
particular subscriber station) for which five subchannels are
D. Packet Dropping Probability allocated and we assume that the average SNR is the same for
It refers to the probability that an incoming packet will be all of these subchannels. Each subchannel has a bandwidth of
dropped due to the unavailability of buffer space. It can be 160 kHz. The length of a subframe for downlink transmission
derived from the average number of dropped packets per is one millisecond, and therefore, the transmission rate in one
frame. Given that there are j packets in the queue and the subchannel with rate ID = 0 (i.e., BPSK modulation and coding
number of packets in the queue increases by v, the number of rate is 1/2) is 80 kbps. We assume that the maximum number
dropped packets is m  ( L  j ) for m  L  j , and zero of packets arriving in one frame for a connection is limited to
30 (i.e., A = 30).
otherwise. The average number of dropped packets per frame is
obtained as follows: For our scheme, the value of the threshold C is varied
according to the evaluation scenarios.

48 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Volume 9 No. 3, March 2011
For performance comparison, we also evaluate the queueing The packet-level performances under different connection
performance in the absence of CAC mechanism. For the case arrival rates are shown in Figures 5 through 8 for average
without CAC, we truncate the maximum number of ongoing number of packets in the queue, packet dropping probability,
connections at 25 (i.e. Ctr  25 ) so that (i, j,Ctr )  2.104,  i, j . queue throughput, and average queueing delay, respectively.
These performance parameters are significantly impacted by
The average duration of a connection is set to ten minutes (i.e., the connection arrival rate. Because the CAC scheme limits the
µ = 10) for all the evaluation scenarios. The queue size is 150 number of ongoing connections, packet-level performances can
packets (i.e., L = 150). The parameters are set as follows: The be maintained at the target level. In this case, the CAC scheme
connection arrival rate is 0.4 connections per minute. Packet results in better packet-level performances compared with
arrival rate per connection is one packet per frame for state 0 of those without CAC scheme.
MMPP process and two packets per frame for state 1 of MMPP
process. Average SNR on each subchannel is 5 dB.
Note that, we vary some of these parameters depending on the
evaluation scenarios whereas the others remain fixed.

B. Performance of CAC policy
We first examine the impact of connection arrival rate on
connection-level performances. Variations in average number
of ongoing connections and connection blocking probability
with connection arrival rate are shown in Figures 3 and 4,
respectively. As expected, when the connection arrival rate
increases, the number of ongoing connections and connection
blocking probability increase.
Figure 5: Average number of packets in queue under different
connection rates.

Figure 3: Average number of ongoing connections under
different connection arrival rates. Figure 6: Packet dropping under different connection arrival
rates.

Figure 4: Connection blocking under different connection Figure 7: Queueing throughput under different connection
arrival rates. arrival rates.

49 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Volume 9 No. 3, March 2011

Figure 8: Average packet delay under different connection Figure 11: Connection blocking probability under different
arrival rates. channel qualities.

Variations in packet dropping probability and average VI. CONCLUSION
packet delay with channel quality are shown in Figures 9 and
10, respectively. As expected, the packet-level performances In this paper, we have addressed the problem of queueing
become better when channel quality becomes better. Also, we theoretic performance modeling and analysis of OFDMA
observe that the connection-level performances for the CAC transmission under admission control. We have considered a
scheme and those without CAC scheme are not impacted by WiMAX system model in which a base station serves multiple
the channel quality when this later becomes better (the subscriber stations and each of the subscriber stations is
connection blocking probability remains constant when the allocated with a certain number of subchannels by the base
channel quality varies) (Figure. 11). station. There are multiple ongoing connections at each
subscriber station.
We have presented a connection admission control
scheme for a multi-channel and multi-user OFDMA network,
in which the concept of guard channel is used to limit the
number of admitted connections to a certain threshold
The connection-level and packet-level performances of
the CAC scheme have been studied based on the queueing
model. The connection arrival is modeled by a Poisson process
and the packet arrival for a connection by a two-state MMPP
process. We have determined analytically and numerically
different performance parameters, such as connection blocking
probability, average number of ongoing connections, average
queue length, packet dropping probability, queue throughput,
Figure 9: Packet dropping probability under different channel and average packet delay.
qualities. Numerical results show that, the performance parameters
of connection-level and packet-level are significantly impacted
by the connection-level rate, the CAC scheme results in better
packet-level performances compared with those without CAC
scheme. The packet-level performances become better when
channel quality becomes better. On the other hand, the
connection-level performances for the CAC scheme and those
without CAC scheme are not impacted by the channel quality.
All the results showed in this paper remain in correlation
with those presented in [10] even if we change here the arrival
packet Poisson process by an MMPP process, which is more
realistic.

REFERENCES
Figure 10: Average packet delay under different channel [1] B. Baynat, S. Doirieux, G. Nogueira, M. Maqbool, and M. Coupechoux,
qualities “An efﬁcient analytical model for wimax networks with multiple trafﬁc
proﬁles,” in Proc. of ACM/IET/ICST IWPAWN, September 2008.

50 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Volume 9 No. 3, March 2011
[2] B. Baynat, G. Nogueira, M. Maqbool, and M. Coupechoux, “An [9] R. Nelson, “Probability, stochastic process, and queueing theory”,
efficient analytical model for the dimensioning of wimax networks,” in Springer-Verlag, third printing, 2000.
Proc. of 8th IFIP-TC6 Networking Conference, May 2009. [10] D. Niyato and E. Hossain, "Connection admission control in OFDMA-
[3] L. B. Le, E. Hossain, and A. S. Alfa, “Queuing analysis for radio link based WiMAX networks: Performance modeling and analysis," invited
level scheduling in a multi-rate TDMA wireless network,” in Proc. chapter in WiMax/MobileFi: Advanced Research and Technology, (Ed.
IEEE GLOBECOM’04, vol. 6, pp. 4061–4065, November–December Y. Xiao), Auerbach Publications, CRC Press, December 2007.
2004. [11] D. Niyato and E. Hossain, “Connection admission control algorithms for
[4] Y. Fang and Y. Zhang, “Call admission control schemes and OFDMA wireless networks,” in Proc. IEEE GLOBECOM’05, St. Louis,
performance analysis in wireless mobile networks,” IEEE Transactions MO, USA, 28 November–2 December 2005.
on Vehicular Technology, vol. 51, no. 2, March 2002, pp. 371–382. [12] D. Pareek, “WiMax: Taking Wireless to the MAX,” Auerbach
[5] D. Hong and S. S. Rappaport, “Traffic model and performance analysis Publishers Inc. June, 2006.
for cellular mobile radio telephone systems with prioritized and [13] R. Ramjee, R. Nagarajan, and D. Towsley, “On optimal call admission
nonprioritized handoff procedures,” IEEE Transactions on Vehicular control in cellular networks,” in Proc. IEEE INFOCOM’96, vol. 1,
Technology, pp. 77–92, August 1986. San Francisco, CA, March 1996, pp. 43–50.
[6] H. Lee and D.-H. Cho, “VoIP capacity analysis in cognitive radio [14] IEEE Std 802.16e-2005 and IEEE Std 802.16-2004/Cor 1-2005, IEEE
system,” IEEE Commun. Lett., vol. 13, no. 6, pp. 393–395, Jun. 2009. Standard for Local and metropolitan area networks-Part 16: Air Interface
[7] H. Lee, Hyu-Dae Kim and Dong-Ho Cho, "Smart Resource Allocation for Fixed and Mobile Broadband Wireless Access Systems, Dec. 7,
Algorithm Considering Voice Activity for VoIP Services in Mobile- 2005.
WiMAX System," IEEE Transactions on Wireless Communications, vol [15] J.-W. So, “Performance analysis of VoIP services in the IEEE 802.16e
8, issue 9, pp. 4688-4697, Sep. 2009. OFDMA system with inband signaling,” IEEE Trans. Veh. Technol.,
[8] H. Lee and Dong-Ho Cho, "Capacity Improvement and Analysis for vol.57, no. 3, pp. 1876–1886, May 2008.
VoIP Service in Cognitive Radio System," IEEE Transactions on
Vehicular Technology, vol 59, issue 4, pp. 1646-1651, May 2010.

51 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Enhancement and Minutiae Extraction of
Touch less Fingerprint Image Using Gabor
and Pyramidal Method

A.John Christopher
Associate Professor, Department of Computer Science, Dr.T.Jebarajan,
S.T. Hindu College, Nagercoil Principal,
V.V. College of Engineering., Tisayanvilai
Abstract - Touch based sensing techniques generate lot of errors
in fingerprint minutiae extraction. The solution for this problem deformation, slippage, smearing or sensor noise. Some of the
is touchless fingerprint technology. They do not receive any touch based are shown in fig.1. A new generation of
contact between the sensor & finger. Although they reduce the touchless live scan devices that generate three various
problems of touch based finger prints, other difficulties explore
such as a view difference problem and a limited usable area due
representation of fingerprint is appearing in the market. This
to perspective distortion. To solve this problem, proposed new sensing technology addresses many of the problems
method for touchless fingerprint image enhancement and stated above [3]. From wear and tear of surface coating, to
minutiae extraction is introduced. Image enhancement is mostly overcome these kinds of problems, a touchless fingerprint
required preprocessing system for finger based biometric sensing technology has been proposed that does not require
system. Normally the touchless device is having a single camera any contact between a sensor and a finger. Thus, the fingers
and two planer mirrors which reflecting side views of a finger. and ridge information cannot be changed or distorted as it
From this we get three images normally frontal, left and right will be free of skin deformation. Also, it can capture
finger. Experimental result shows that the enhanced images fingerprint images consistently because it is not affected by
increase the biometric accuracy.
different skin conditions or latent fingerprints.
Index Terms - pyramidal method, Gabor, touchless fingerprint,
thinning, normalization, finger enhancement, adaptive histogram.

I ‐ INTRODUCTION

A fingerprint is composed of ridges and valleys.
Ridges have various kinds of discontinuity such as ridge
bifurification, ridge endings, short ridges, islands and ridge
cross over’s. Among this discontinuity, ridge bifurification
and ridge ending are commonly used in fingerprint
identification/verification system and are called minutiae
[1].For the processing of fingerprint images, two stages are of
pivotal importance for the success of biometric
reorganization: image enhancement and minutiae extraction.
The traditional fingerprint processing technologies are
applied immediately after sensing. But a better thing is an
optional image enhancement in fingerprint images. In
Fig. 1: Distorted images acquired from a touch-based sensor.
realistic scenarios though the quality of a fingerprint image
may suffer from various impairments, caused by scores, cuts,
Recently, several companies and research groups have
moist or dry skin, sensor noise, blur, wrong handling of
developed touchless fingerprint sensors and recognition
sensor, weak ridge and valley pattern of the given fingerprint,
systems [4]–[6]. TST Group developed a touchless imaging
etc. The task of the fingerprint enhancement is to counteract
sensor (BiRD III) which uses a complementary metal–
the aforesaid quality impairments and to reconstruct the
organic–semiconductor (CMOS) camera, and red and green
actual fingerprint pattern as trace to it original as possible. [2]
light sources to acquire fingerprint images [4]. Song et al. [5]
Fingerprints are traditionally captured based on contact of the
proposed a sensing system with a single charged-coupled
finger on paper or a platen. This often results in partial or
device (CCD) camera and double ring-type blue illuminators
degraded images due to improper finger placement, skin
to capture high contrast images. Also, Mitsubishi Electric
Corporation proposed another touchless approach

52 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

transmitting the light through the finger [6], acquiring can capture three different views of a fingerprint using only
fingerprint patterns under the surface of skin using light with one Camera and also avoid the synchronization problem
a wavelength of 660 nm. However, such sensing systems [4]– existing in multiple camera-based systems. In addition, to
[6] have an inherent problem as they use only a single obtain high-quality fingerprint images, we need to consider
capturing device, such as CMOS or CCD cameras. when several optical components in order to design the device.
capturing an image using a single camera, the geometrical
resolution of the fingerprint image decreases from the
fingerprint center towards the side area [7]. Therefore, false
features may be obtained in the side area and it reduces the
valid and useful region for authentication. Moreover, if there
is a view difference between images due to finger rolling, it
reduces the common area between fingerprints and degrades
system performance. To solve this problem, 3-D touchless
sensing systems using more than one view have been
explored [8]–[11]. TBS [8] used five cameras placed around
a finger to capture nail-to-nail fingerprint images and Fig. 2: Proposed device.
generated a 3-D fingerprint image using the shape-from- (a) Prototype of the device. (b) Schematic view of the device.
silhouette method. They then unwrapped the 3-D finger
image onto a 2-D image by using parametric and The specifications of the optical components are as follows:
nonparametric models to make rolled-equivalent images [9]. 1) Camera and lens: We use a 1/3-in progressive scan
Fatehpuria et al. [10] proposed a 3-D touchless device using type CCD with 1024 x 768 active pixels, where the
multiple cameras and structured light illumination (SLI). The pixel size is 4.65 x 4.65 m. This camera offers a
structured light patterns are projected onto a finger to obtain sufficient frame rate of 29 Hz, thus avoiding image
its 3-D shape information and 2-D unfolded images are blurring caused by typical finger motion. Also, we
generated by applying “Springs algorithm” and some post use simple equations [see (1) and (2)] to design an
processing steps. Also, the Hand Shot ID system was adequate lens for our system.
developed to acquire a 3-D shape of a hand with fingers by
stitching images from 36 cameras [11]. Although all these
q
methods attempted to solve the problems in touch-based M = (1)
sensors and acquire expanded fingerprint images with less p
skin deformation, they did not raise much interest in the
1 1 1
market because of much higher costs compared to = + (2)
conventional touch-based sensors. Considering the above f p q
observations, we adopt a new touchless sensing scheme using
a single camera and a set of mirrors. The mirrors work as Where f is the lens focal length, p and q are the lens-
virtual cameras, thus enabling the capture of an expanded to-object and lens-to-image distances, respectively,
view of a fingerprint at one time without using multiple and M is the optical magnification. Normally, the
cameras. The device consists of a single camera, two planar required image resolution for touch-based sensors is
mirrors, light-emitting diode (LED)-based illuminators, and a 500 dpi. Therefore, to ensure a 500-dpi spatial
lens. Two planar mirrors are used to reflect the left and right resolution in the fingerprint area and to cover three
side view of a finger. In this paper, we proposed a new view fingerprints, the optical magnification
method to enhance the touchless finger print and to extract parameter M, the lens to image distance, and field of
the minutiae data. view (FOV) are determined as 0.1, 170 mm, and 50
II – SYSTEM DESIGN x 38 mm, respectively. By doing this, we can
capture three view images with 500-dpi resolution at
To overcome the view difference problem and the one time. Also, the depth of field (DOF) of the lens
limitation of a single view, some touchless fingerprinting ranges from -2.6 to +2.6 mm at a given working
systems capture several different views of a finger by using distance and it normally covers the half depth of a
multiple cameras. However, using multiple cameras increases finger.
the cost and size of a system. Thus, we adopt a new sensing
system which captures three different views (frontal, right, 2) Illumination: Considering the reflectance of human
and left) at one time by using a single camera and two planar skin to various light sources, we used ring-shaped
mirrors. Figs. 2(a) and (b) show the prototype and schematic white LED illuminators and a band pass filter which
view of the device. As shown in Fig. 2, two mirrors are can transmit green light to enhance the ridge-to-
placed next to the finger and reflect the right and left side valley contrast. Also, the illuminators are placed
views of the finger. Then, the frontal view and two mirror- perpendicular to the finger to remove the shadowing
reflected views are captured by a single camera effect. Diffusers are used to illuminate a finger
simultaneously. A mirror-reflected image is regarded as the uniformly.
“flipped” image taken by a virtual camera placed at a
different direction compared to the real one. Therefore, we

53 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

can be used to fill "holes" of a size equal to or
smaller than the structuring element. Used with
binary images, where each pixel is either 1 or 0,
dilation is similar to convolution. Over each pixel of
the image, the origin of the structuring element is
Foreground separation overlaid. If the image pixel is nonzero, each pixel of
the structuring element is added to the result using
the "or" operator. Used with greyscale images,
Normalization which are always converted to byte type, the
DILATE function is accomplished by taking the
maximum of a set of sums. It can be used to
Gabor filtering conveniently implement the neighbourhood
maximum operator with the shape of the
neighbourhood given by the structuring element.
Used with greyscale images, which are always
Pyramidal method
converted to byte type, the ERODE function is
accomplished by taking the minimum of a set of
differences. It can be used to conveniently
Thinning implement the neighbourhood minimum operator
with the shape of the neighbourhood given by the
structuring element.
Minutiae extraction B) Normalisation
The process of removing the effects of the sensor
noise and gray-level background due to finger
pressure differences. The objective of this stage is
Fig. 3: Overall flowchart of the proposed method decrease the dynamic range with gray scale between
ridges and valleys of the image. Normalization
3) Mirror: Two planar mirrors are positioned next to factor is calculated according to the mean and the
the left and right side of the finger and the mirror variance of the image. Each and every pixel in the
size is determined to cover the maximum thumb fingerprint image has to be processed to find the
size. To provide enough overlapping area between median value. The average value of all the pixels is
frontal- and side-view images, the angles of the calculated i.e, the median value. By comparing the
mirrors are determined 15 empirically. Also, the median value with the current pixel the replacement
mirrors can be used as pegs to place a user’s finger can be performed.
firmly on the device. Normalization facilitates have the subsequent
processing steps.
III – PROPOSED METHOD Let G (i, j) denote the normalized gray-level value at
In this section, we explain the Enhancement method for pixel (i, j). The normalized image is defined as
synthesizing an expanded fingerprint image from frontal- and follows:
side-view images. The overall scheme of the method is
presented in Fig. 3 The method is mainly composed of six
stages (foreground separation, normalisation, Gabor filtering,
pyramidal method, thinning, minutiae extraction). In (3)
foreground separation we will do the morphological
operation, in normalisation we pre-process the image etc.
A) Foreground separation Where, M 0 and VAR0 denote the desired
Using morphological operation we use the erosion mean and variance value, respectively.
followed by dilation, this can be done up to required
Most fingerprint images on a live-scan input device
time. Mathematical morphology is a method of
processing digital images on the basis of shape. A are usually of poor quality. The fingerprint image is
discussion of this topic is beyond the scope of this smoothed with an average or median filter.
manual. A suggested reference is: Haralick, C) Gabor filtering
Sternberg, and Zhuang, "Image Analysis Using A Gabor filter is a linear filter used in image
Mathematical Morphology," IEEE Transactions on processing for edge detection. Frequency and
Pattern Analysis and Machine Intelligence, Vol. orientation representations of Gabor filter are similar
PAMI-9, No. 4, July, 1987, pp. 532-550. Much of to those of human visual system, and it has been
this discussion is taken from that article. Briefly, the found to be particularly appropriate for texture
DILATE function returns the dilation of image by representation and discrimination. In the spatial
the structuring element Structure. This operator is domain, a 2D Gabor filter is a Gaussian kernel
commonly known as "fill", "expand", or "grow." It function modulated by a sinusoidal plane wave. The

54 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Gabor filters are self-similar - all filters can be Reduce the image size by a factor k for three times.
generated from one mother wavelet by dilation and This is also outlined on the upper left hand side of Table
rotation. Its impulse response is defined by a 1. To create images containing only band limited signals
harmonic function multiplied by a Gaussian of the original image, we expand the three images by
function. Because of the multiplication-convolution factor and subtract each of them from the next lower
property (Convolution theorem), the Fourier level.
transform of a Gabor filter's impulse response is the E) Thinning
convolution of the Fourier transform of the The THIN function returns the "skeleton" of a bi-
harmonic function and the Fourier transform of the level image. The skeleton of an object in an image is
Gaussian function. a set of lines that reflect the shape of the object. The
g ( x, y; λ,θ ,ϕ,σ , γ ) set of skeletal pixels can be considered to be the
medial axis of the object. For a much more
extensive discussion of skeletons and thinning
algorithms, see Algorithms for Graphics and Image
Processing, Theo Pavlidis, Computer Science Press,
1982. The THIN function is adapted from Algorithm
9.1 (the classical thinning algorithm).On input, the
bi-level image is a rectangular array in which pixels
(4) that compose the object have a nonzero value. All
other pixels are zero. The result is a byte type image
in which skeletal pixels are set to 2 and all other
Where x ' = x cos θ + y sin θ and pixels are zero.
F) Minutiae extraction
A feature extractor finds the ridge endings and ridge
y = − x sin θ + y cos θ
'
bifurcations from the input fingerprint images. If
ridges can be perfectly located in an input
In this equation, λ represents the wavelength of the fingerprint image, then minutiae extraction is just a
cosine factor, θ represents the orientation of the trivial task of extracting singular points in a thinned
normal to the parallel stripes of a Gabor function, φ ridge map. However, in practice, it is not always
is the phase offset, σ is the sigma of the Gaussian possible to obtain a perfect ridge map. The
envelope and γ is the spatial aspect ratio, and performance of currently available minutiae
specifies the ellipticity of the support of the Gabor extraction algorithms depends heavily on the quality
function. of the input fingerprint images. Due to a number of
factors (aberrant formations of epidermal ridges of
D) Pyramidal method fingerprints, postnatal marks, occupational marks,
Pyramid decomposition requires resizing problems with acquisition devices, etc.), fingerprint
(scaling, or other geometric transformation). To images may not always have well-defined ridge
create our Gaussian and Laplacian like pyramids, we structures. A reliable minutiae extraction algorithm
define the reduce(I,K) and expand(I,K) operations, is critical to the performance of an automatic
which decrease and increase an image in size by the identity authentication system using fingerprints.
factor K, respectively. During reduce, the image is
initially low-pass filtered to prevent aliasing using a
Gaussian kernel.2. The latter’s standard deviation
depends on the resizing factor, which here follows
the lower bound approximation of the corresponding
ideal low-pass filter . We initially
reduce the original fingerprint image FP by a factor
of in order to exclude the highest
frequencies. In a further step, we Fig. 4: Types of Ridge Patterns

Table - 1
Pyramidal building process

a) Pyramidal decomposition
Gaussian-like Laplacian-like

G1=reduce(fp,k0) L1=g1-expand(g2,k)
Fig. 5: Minutiae points
G2=reduce(g1.k) L2=g2-expand(g3,k)

55 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Minutiae are extracted from the thinned image by checking methods compare the foreground size of the fingers.
using the Crossing Number algorithm. Here foreground means the good quality regions of the finger
print. The foreground size measures are tabulated as follows:

(5)

Where Pi 0 or 1 in the 3*3 Neighbor of P

Characteristic of CN
CN Character
Fig. 8: Minutiae
0 Isolated point
Table - 2
2 End point Average increasing rate of Foreground size in terms of each measurement

4 Bifurcation point Quality measurement Average increase rate of
foreground size
Standard deviation [12] 28.65%
IV – EXPERIMENTAL RESULTS
For the experimental results we acquired 100 set of finger Coherence [13] 33.72%
print images, each set contain frontal, left and right view
images. One of the used images set is shown in the Fig: 6 and Gradient – based 30.81%
the enhanced image is also shown in the Fig: 7. The minutiae method [14]
extraction results also expressed in Fig: 8. The most definite
indicator of touchless image quality is the number of true However we can expect that our enhanced image can be
minutiae additionally extracted. making high performance when view difference image are
matched. The Table-2 shows the result of our enhanced
image.
V – CONCLUSIONS AND FUTURE WORK
This paper proposes a new method for touchless fingerprint
sensing images. To get the better minutiae extraction, the
three fingerprints (frontal, left, right) are enhanced using
Gabor and pyramidal method. For experimental results, the
enhanced fingerprints are having better enhanced ridges and
the valleys. Also minutiae extraction is handled. The results
are analysed and described in tables and graph format. In this
Fig. 6: Input images paper we limits the research work up to minutiae extraction,
this research can be continued on mosaicing of the three
enhanced images. Feature work can be done on the same
concept. According to the result, it is concluded that the
proposed system generate better enhancement on touchless
fingerprint than the existing methods.

REFERENCES
[1] D. Lee, K. Choi, and J. Kim, “A robust fingerprint matching
algorithm using local alignment,” in Proc. 16th Int. Conf. Pattern
Recognition, 2002, vol. 3, pp. 803–806.
[2] Hartwig Fronthaler, Klaus Kollreider, and Josef Bigun ,Local
Features for Enhancement and Minutiae Extraction in
Fingerprints, IEEE Transactions On Image Processing, VOL. 17,
NO. 3, MARCH 2008
Fig. 7: Enhanced images [3] Yi Chen1, Geppy Parziale2, Eva Diaz-Santana2, and Anil K Jain,
“3d Touchless Fingerprints: Compatibility With Legacy Rolled
Human experts prove that the more true minutiae extracted Images” Michigan State University Department of Computer
from the enhanced image. The touchless fingers are better Science and Engineering, 2006 Biometrics Symposium,
[4] TST Group Aug. 03, 2009 [Online]. Available: http://www.tst-
than the conventional touch based fingers, that conclusion biometrics.com
can be deviate from the results. The finger print quality

56 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

[5] Y. Song, C. Lee, and J. Kim, “A new scheme for touchless
fingerprint recognition system,” in Proc. Int. Symp. Intelligent
Signal Processing and Communication Systems, 2004, pp. 524–
527.
[6] Mitsubishi Touchless Fingerprint Sensor Aug. 03,
2009[Online].Available: http://global.mitsubishielectric.com
[7] N. K. Ratha and V. Govindaraju, Advances in Biometrics:
Sensors, Algorithms and Systems. New York: Springer, 2008
[8] TBS Touchless Fingerprint Imaging Aug. 03, 2009
[Online].Available: http://www.tbsinc.com/
[9] Y. Chen, G. Parziale, E. Diaz-Santana, and A. K. Jain, “3D
touchless fingerprints: Compatibility with legacy rolled images,”
in Proc. Biometric Consortium Conf., Baltimore, MD, 2006.
[10] A. Fatehpuria, D. L. Lau, and L. G. Hassebrook, “Acquiring a 2-
D rolled equivalent fingerprint image from a non-contact 3-D
finger,” in SPIE Defense and Security Symp. Biometric
Technology for Human Identification III, Orlando, FL, 2006, vol.
6202, pp. 62020C-1–62020C-8.
[11] Aug. 03, 2009 [Online]. Available:
http://privacy.cs.cmu.edu/dataprivacy/ projects/handshot
/index.html
[12] L. Hong, Y.Wan, and A. K. Jain, “Fingerprint image
enhancement: Algorithm and performance evaluation,” IEEE
Trans. Pattern Anal. Mach Intell., vol. 20, no. 8, pp. 777–789,
Aug. 1998.
[13] E. Lim, X. Jiang, and W. Yau, “Fingerprint quality and validity
analysis,”in IEEE Int. Conf. Image Processing (ICIP), Sep. 2002,
vol. 1,pp. 469–472.
[14] S. Lee, H. Choi, and J. Kim, “Fingerprint quality index using
gradientcomponents,” IEEE Trans. Inf. Forensics Security, vol. 3,
no. 4, pp.792–800, Dec. 2008.
http://www.youtube.com/watch?v=5ntH8s03ujk&feature=relat
ed

57 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Automatic parsing For Arabic sentences

Zainab Ali Khalaf* Dr. Tan Tien Ping

School of computer science School of computer science
Universiti Sains Malaysia (USM) Universiti Sains Malaysia (USM)
Penang, Malaysia Penang, Malaysia
E-mail: zak10_com026@student.usm.my E-mail: tienping@cs.usm.my
*(Ass. Prof. In Computer Science Dept.,
Basra University, Iraq)

Abstract__The designed system is a parser for Arabic
sentences using syntactic and semantic relations The proposed system aims to use these properties
between deep and surface structures. The system to parse Arabic sentences depending on the position
depends on implementation of Case theory of Fillmore. of the words in the sentence and the functional
meaning of them.
The parsing algorithm starts analyzing the input
sentence to check its syntax, semantic and spelling using
Arabic transformation rules proposed in Al_Khouly to
gain semantic strength. The proposed system depends II. SYSTEM COMPONENTS
on the effective elements represented by the verb of the
sentence .This element is used to control the parsing
operation. The syntactical properties of any natural
The proposed system permits as input different
language are formally described by the use of what
surface structures of Arabic sentences to produce
automatic parsing forms for these input sentences. Chomsky calls production systems. A formal system
generally depends on three types of data [2,3,6]:
Keywords__Artificial intelligence; natural language
processing; transformation rules; deep structure and A. Data of vocabulary lexicon
surface structure; parsing Arabic sentences .

The lexicon plays an important role in any NLP
system. It is a huge data base of variable entries
I. INTRODUCTION describing the meaning of words in synonymy (and
antinomy) contextual fashion [3,6]. The implemented
lexicon consists of entries saved as a rule ( Entrance
Arabic language is a parsing language . Parsing [ Word , Features ] ).
means the relation among the words in the sentence.
The most important component is the verb which acts • The Entrance is one of the following indicators :-
as the basic unit to control the rules of choosing other Verb , Noun , Preposition , Determinate , Assistant
elements. Although Arabic sentences have different and Negation.
structures , but it is recognized as a ( verb , subject ,
object ) language. The subject or the object may be The Word is a string index for the lexicon entry.
precede the verb in the Arabic sentences according to
the pragmatic necessity [1,3,4].
• The Features is a list of structured integers coded
to hold the syntactical and semantic information of
Arabic Syntactic facilitates the flexibility of the the word. Each coded integer, written as [Fp],
deep structure and the surface structure of sentence to consists of two parts F and p. The [p] part is either 1
be connected together strongly. This propriety helps or 0 depending on whether the feature [F] exists or
Arabic language accept for automatic processing not. The [F] part is the feature code.
[4,5].

58 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

B. Data of syntactical rules
The presence of the verb is necessary and obligatory,
whereas the presence of other elements is optional
These rules are formalized to describe the and dependent on the verb rules [1,4].
language in order to relate each one deep structure
into so many corresponding surface structures of the
same meaning. These rules are actually inductive and
sequential. Some are obligatory and others are III. DENESIGD SYSTEM STRUCTURE
optional rules. From the optional rules, one can obtain
various surface structures that act as contextual The designed system has many stages : Figure (1)
linguistics. The transformations are mainly operations acts flowchart of these stages which are described
that are addition, deletion, moving forward, moving below :
backward and some other secondary operations. These
operations are, in general, not performed at random,
but are governed and selected according to a set of A. Input sentence stage
conditions and rules of structure description. These
operations will generate all surface structures
emerging from that one deep structure. The function of this stage is to input Arabic sentence
from the keyboard to the computer , this sentence
ended by dot or semicolon or space character .
C. Data of syntactic structure

These data are rules described in BNF for
Arabic language , and acts as constraints and controls B. Segmentation stage
to form the sentences of Arabic language. The most
important component, as Fillmore and Shank
recognized, is the verb element which acts as the The function of this stage is to segment the input
basic unit that controls rules of choosing other sentence into words depended on space character
elements. The dependent phrase structure rules used (free number of space characters).
are the following :-
C. Lexicon search stage
<Sentence> ::= <Modality> + < Auxiliary > + <
Proposition >
<Sentence> ::= < Auxiliary > + < Proposition > < The function of this stage is to search for all sentence
Modality > ::= < External Condition > + < External words in the lexicon . If the word is not found in the
Adverb > lexicon, the program gives spelling error message
<Proposition > ::= < Verb > + < Theme > + < Indirect and stop .
Object > + < Place > + < Tool > + < Agent >
< Theme > ::= < Noun Phrase > D. Syntactical analysis stage
< Agent > ::= < Noun Phrase >
< Tool > ::= < Noun Phrase >
< Place > ::= < Noun Phrase > The function of this stage is to ensure and govern the
< Indirect Object > ::= < Noun Phrase > correctness of input sentence from its syntactical side
<Noun phrase> ::=<Proposition> + <determinate > + < . If the processing found errors , the program gives
Noun > syntactical error massage .
< Noun Phrase > ::= < Proposition >+ < Noun>
< External Condition > ::= semi statements used to
combine two sentences such as ( in spite of E. Semantic analysis stage
) or ( moreover ) etc.
< External Adverb > ::= <Time Adverb>+<
Interrogative Words> +<Negation Words> The function of this stage is to ensure and govern the
< Auxiliary > ::= lexical words such as ( ) or ( correctness of input sentence from its harmony, its
) etc. vocabulary and correctness of its meaning . If the
< Verb > ::= A dictionary verb such as ( write ) sentence is not correct in its meaning, the program
etc. gives semantic error massage .
< Noun > ::= A dictionary noun such as ( boy )
etc.

59 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

F. Generative deep structure stage

IV. EXAMPLES
Transformational operations will carry out , and try
to compile the addition, deletion, replacement and
other operations to obtain on the sentence structure
which acts as the deep structure . For example we want to know the parsing of the
following sentences. Figure (2) depicted this
mechanism :-
G. Parsing stage
A. Example 1
The function of this stage is to parse sentence which
depends on its effective element and its position in
structure phrase . This stage has many Arabic The system prints the following parsing :
language rules which control the parsing operations . :
. . :
Here an examples of sentences that the system can :
parse its :-
.
B. Example 2
.1
.
.2 The system prints the following parsing :

.3 :
.
.4 . :
. :
.5
C. Example 3
.6
.
.7 The system prints the following parsing :

.8
.9 :
.
.10 :
.
.11 :
.
.12 D. Example 4

.13 .
The system prints the following parsing :
.14
.15 . :
. :
.16 :
.17 .
E. Example 5
.
The system prints the following parsing :

60 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

References
. :
: [1] Abo-Arafah .A. , "A grammar for the Arabic language suitable
. for machine parsing and automatic text generation ", PH.D. thesis ,
: Illinois of technology , Chicago , USA,1995 .
.
. :
[2] Ali .N. , “Arabic language and Computer” , "Al-Tareeb
Publishing House, Cairo, Egypt, 1988.

[3] Al-Khouly, M. , “ Transformation rules for Arabic language”,
Al- Riyadh, 1981.
Conclusions
[4] Al-Shalabi .R. , Evens .M ." A Computational Morphology
System For Arabic " , Dept. Of Computer Science and applied
Mathematics , Illinos Institute Of Technology , Chicago , USA ,
The present research ends up with the following W.D.
conclusions :-
[5] Gheith .M. , Mashour .M . " A Computer Based System For
1. The verb is the main component which controls understanding Arabic language ", Computer Science Department
all other component appearing with it . From this Inst. Of Statistical Study & Research , Cairo University , Egypt
point, we consider all deep structures as containing ,W.D.
the verb in its structure .
[6] Khalaf .Z. , “Computerized Implementation For Processing
Arabic Sentences By Interpretation Synonymy Relationships” ,
2. The word meaning depends on the essential M.Sc. thesis, Basra University, Iraq, 2001.
effective element ( the deep element ) .

3. The lexicon plays the essential element to
provide any system by vocabulary and its features .
By these features, we can control the different
processing levels of syntax and semantics .

4. The absence of vowelization might bring some
ambiguities to sentence understanding. However the
transformation rules are used to remedy these
ambiguities in an explicit and easy way, as in the
following sentences which show where, in all the
sentences, the man is the subject and the lion is object
.

Acknowledgment

I would like to express my sincere appreciation to
TWAS organization , USM university for their
encouragement and continuous financial support
through the providing PHD fellowship. In addition we
would like to thank school of computer science for
their encouragement and motivation of international
students in the faculty.

61 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

User Interface

Input Stage

Segmentation
Lexicon
Stage
Lexical Rules

Spelling
Lexicon Stage
Error

Initial Descriptive
Structure

Syntactical
Transformational Rules Errors
Transformational Rules

Semantic
Stage Deep Structure

Semantic Parsing stage
Error

User Interface

Figure (1) acts flowchart of
Parsing operations

62 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Surface structure

Transformation Rules

An agent ( ) used a tool ( ) to
perform the verb ( ) to get the object
( )
Deep structure

Verb ( ) , Subject ( ) , Object ( ) , Tool ( )

Sentence structure

Parsing Stage

:
. :
. :
. :

Figure (2) acts the mechanism to Parse
Arabic sentence

63 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011

Amelioration of Walsh-Hadamard Texture Patterns
based Image Retrieval using HSV Color Space
Dr. H.B.Kekre1, Sudeep D. Thepade2, Varun K. Banura3
1
Senior Professor, 2Ph.D.Research Scholar & Associate Professor, 3B.Tech (CE) Student
Computer Engineering Department, MPSTME,
SVKM‟s NMIMS (Deemed-to-be University), Mumbai, India
1
hbkekre@yahoo.com, 2sudeepthepade@gmail.com,3varunkbanura@gmail.com

Abstract— The theme of the work presented here is amelioration A. Content Based Image Retrieval
of Walsh-Hadamard texture pattern based image retrieval using
HSV color space. Different texture patterns namely ‘4-pattern’, For the first time Kato et.al. [7] described the experiments
‘16-pattern’, ‘64-pattern’ are generated using Walsh-Hadamard of automatic retrieval of images from a database by colour and
transform matrix and then compared with the bitmap of an shape feature using the terminology content based image
image in HSV color space to generate the feature vector as the retrieval (CBIR). The typical CBIR system performs two major
matching number of ones and minus ones per texture pattern. tasks [19,20] as feature extraction (FE), where a set of features
The proposed content based image retrieval (CBIR) techniques called feature vector is generated to accurately represent the
are tested on a generic image database having 1000 images content of each image in the database and similarity
spread across 11 categories. For each proposed CBIR technique measurement (SM), where a distance between the query image
55 queries (randomly selected 5 per category) are fired on the and each image in the database using their feature vectors is
image database. To compare the performance of image retrieval used to retrieve the top “closest” images [19,20,29].
techniques average precision and recall of all the queries per
image retrieval technique are computed. The results have shown For feature extraction in CBIR there are mainly two
improved performance (higher precision and recall values of approaches [8] feature extraction in spatial domain and feature
crossover points) with the proposed methods compared to the extraction in transform domain. The feature extraction in
texture based image retrieval in RGB color space. Further the spatial domain includes the CBIR techniques based on
performance of proposed image retrieval methods is enhanced histograms [8], BTC [4,5,19], VQ [24,28,29]. The transform
using even image part. The proposed CBIR methods do not give domain methods are widely used in image compression, as they
better performance with image bitmaps generated using tiling in give high energy compaction in transformed image [20,27]. So
HSV color space. In the discussed image retrieval methods, the it is obvious to use images in transformed domain for feature
combination of original and even image part for 16-pattern extraction in CBIR [26]. But taking transform of image is time
texture with image bitmaps in HSV color space gives the highest consuming. Reducing the size of feature vector using pure
crossover point of precision and recall indicating better image pixel data in spatial domain and getting the improvement
performance. in performance of image retrieval is shown in [1,2,3]. But the
problem of feature vector size still being dependent on image
Keywords- CBIR, Walsh-Hadamard transform, Texture,
Pattern, Bitmap, HSV color space
size persists in [1,2,3]. Here the query execution time is further
reduced by decreasing the feature vector size further and
I. INTRODUCTION making it independent of image size. Many current CBIR
systems use the Euclidean distance [4-6,11-17] on the extracted
Today the information technology experts are facing feature set as a similarity measure. The Direct Euclidian
technical challenges to store/transmit and index/manage image Distance between image P and query image Q can be given as
data effectively to make easy access to the image collections of equation 1, where Vpi and Vqi are the feature vectors of image
tremendous size being generated due to large numbers of
P and Query image Q respectively with size „n‟.
images generated from a variety of sources (digital camera,
digital video, scanner, the internet etc.). The storage and n
transmission is taken care of by image compression [4,7,8]. ED   (Vpi  Vqi )
i 1
2
(1)
The image indexing is studied in the perspective of image
database [5,9,10,13,14] as one of the promising and important
research area for researchers from disciplines like computer II. TEXTURE PATTERNS USING WALSH-HADAMARD
vision, image processing and database areas. The hunger of TRANSFORM MATRIX
superior and quicker image retrieval techniques is increasing
day by day. The significant applications for CBIR technology Walsh transform matrix [21,22,26] is defined as a set of N
could be listed as art galleries [15,17], museums, archaeology rows, denoted Wj, for j = 0, 1, .... , N - 1, which have the
[6], architecture design [11,16], geographic information following properties:
systems [8], weather forecast [8,25], medical imaging [8,21],
trademark databases [24,26], criminal investigations [27,28],  Wj takes on the values +1 and -1.
image search on the Internet [12,22,23]. The paper attempts to  Wj[0] = 1 for all j.
provide better and faster image retrieval techniques.

64 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
 WjxWkT=0, for j not equal to k and WjxWkT =N,
for j=k.
 Wj has exactly j zero crossings, for j = 0, 1, .... , N-1.
 Each row Wj is even or odd with respect to its
midpoint

Walsh transform matrix is defined using a Hadamard
matrix of order N. The Walsh transform matrix row is the row
of the Hadamard matrix specified by the Walsh code index,
which must be an integer in the range [0, ..., N - 1]. For the
Walsh code index equal to an integer j, the respective
Hadamard output code has exactly j zero crossings, for j = 0,
1, ... , N - 1.
Using the Walsh-Hadamard transform assorted texture
patterns namely 4-pattern, 16-pattern and 64-pattern are 2(a). 4x4 Walsh-Hadamard transform matrix
generated. To generate N2 texture patterns, each column of the
Walsh-Hadamard matrix of size NxN is multiplied with every
element of all possible columns of the same matrix (one
column at a time to get one pattern). The texture patterns
obtained are orthogonal in nature.
Figure 1(a) shows a 2X2 Walsh-Hadamard matrix. The
four texture patterns generated using this matrix are shown in
figure 1(b). Similarly figure 2(b) shows first four texture
patterns (out of total 16) generated using 4X4 Walsh-
Hadamard matrix shown in figure 2(a).

2(b). First four of the sixteen Walsh-Hadamard texture
patterns (16-pattern)
Figure 2. Generation of sixteen Walsh-Hadamard texture
1(a). 2x2 Walsh-Hadamard transform matrix patterns (16-pattern)
III. GENERATION OF IMAGE BITMAPS
Image bitmaps of colour image are generated using three
independent red (R), green (G) and blue (B) components of
image to calculate three different thresholds. Let
X={R(i,j),G(i,j),B(i,j)} where i=1,2,….m and j=1,2,….,n; be
an m×n color image in RGB space. Let the thresholds be TR,
TG and TB, which could be computed as per the equations
given below as 2, 3 & 4.

1 m n
TR   R(i, j)
m * n i 1 j 1
(2)

1 m n
1(b). Four Walsh-Hadamard texture patterns (4-pattern)
TG   G(i, j)
m * n i 1 j 1
(3)

Figure 1. Generation of four Walsh-Hadamard texture patterns
(4-pattern)

65 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011

1 m n 3
TB   B(i, j)
m * n i 1 j 1
(4) S  1
RG B
[min( R, G, B)] (9)

Here three binary bitmaps are computed as BMr, BMg and
1
BMb. If a pixel in each component (R, G, and B) is greater V  ( R  G  B) (10)
than or equal to the respective threshold, the corresponding 3
pixel position of the bitmap will have a value of 1 otherwise it
will have a value of -1.
V. PROPOSED CBIR METHODS
After generating bitmap of the image in HSV color space,
 1, if .R(i, j )  TR
 to generate feature vectors the bitmap of each image is
BMr(i, j )   (5) compared with the generated texture patterns to find matching
 1, if .R(i, j )  TR
 number of ones and minus ones. The size of the feature vector
of the image is given by equation 11.
 1, if .G(i, j )  TG Feature vector size=2*3*(no. of considered texture-pattern) (11)

BMg(i, j )   (6) Using three assorted texture pattern set along with original
 1, if .G(i, j )  TG
 and original-even image in HSV color space, total six novel
feature vector generation methods can be used resulting into
six new image retrieval techniques. Walsh-Hadamard texture
 1, if .B(i, j )  TB pattern [30,31,32] based image retrieval techniques in RGB

BMb(i, j )   (7) color space are considered to compare the performance of
 1, if .B(i, j )  TB proposed CBIR techniques. In the proposed CBIR techniques

the combination of original and even part of images give better
To generate tiled bitmaps, the image is divided into four non- results than original image alone [1,2]. The proposed CBIR
overlapping equal quadrants and the average of each quadrant techniques do not give good performance with bitmaps
is considered to generate the respective tile of the image generated using tiling [30]. The main advantage of proposed
bitmap. CBIR methods is reduced time complexity for query execution
due to reduced size of feature vector resulting into faster
IV. COLOR SPACE [33] image retrieval with better performance. Also the feature
Color model is an abstract mathematical model describing vector size is independent of image size in proposed CBIR
the way colors can be represented as tuples of numbers, methods.
typically as three or four values or color components. Color Table 1. Feature vector size of discussed image retrieval techniques
space is set of colors where the color model is associated with
a precise description of how the components are to be Feature
interpreted. CBIR vector size
Technique for Binary
A. RGB Color Space Image Maps
RGB uses additive color mixing, because it describes what 8
4-Pattern
kind of light needs to be emitted to produce a given color.
RGB stores individual values for red, green and blue. 16-Pattern 32

B. HSV Color Space 64-Pattern 128
The HSV stands for the Hue, Saturation, and Value based
on the artists (Tint, Shade, and Tone). The Value represents
intensity of a colour, which is decoupled from the colour VI. IMPLEMENTATION
information in the represented image. The Hue and Saturation
components are intimately related to the way human eye The implementation of the discussed CBIR techniques is
perceives colour resulting in image processing algorithms with done in MATLAB 7.0 using a computer with Intel Core 2 Duo
physiological basis. Conversion formula from RGB to HSV is Processor T8100 (2.1GHz) and 2 GB RAM.
given by equations 8, 9 and 10. The CBIR techniques are tested on the Wang image
database [18] of 1000 variable size images spread across 11


1
( R  G)  ( R  B)  categories of human being, animals, natural scenery and
1
H  cos  2
 (8) manmade things, etc. The categories and distribution of the
 ( R  G )  ( R  B)(G  B) 
2
images is shown in table 2.
 

66 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
To analyze the effectiveness of proposed CBIR techniques, Figure 3 shows the performance comparison of proposed
the crossover points of average precision and recall values of CBIR methods in RGB and HSV color space. It is observed
the 55 queries (randomly selected 5 from each image that the performance of the Walsh-Hadamard texture pattern
category) have been used as statistical comparison parameters based image retrieval [30] is improved in HSV color space as
[4,5]. These precision and recall have been defined in the compared to RGB color space. Also the performance of the
equations 12 and 13. texture pattern based image retrieval increases with increase in
number of generated texture patterns up to a certain level (16-
Number _ of _ relevant _ images _ retrieved
Pr ecision  (12) pattern) and beyond this level the results start deteriorating.
Total _ number _ of _ images _ retrieved The „16-pattern‟ texture based image retrieval with the
Number _ of _ relevant _ images _ retrieved combination of original and even image in HSV color space
Re call  (13)
Total _ number _ of _ relevent _ images _ in _ database has the highest crossover point indicating better performance.
Moreover as the number of texture patterns generated is
increased the size of the feature vector also increases thus
Table 2. Image Database: Category-wise Distribution increasing the time complexity for query execution.
Category Monuments Beaches Buses
No. of
99 99 99
Images
Category Dinosaurs Sunrise Tribes
No. of
99 61 85
Images
Category Elephants Horses Roses
No. of
99 99 99
Images
Category Airplanes Mountains
No. of
100 61
Images

Figure 4. Performance comparison of the proposed CBIR methods with the
VII. RESULTS AND DISCUSSIONS combinationof original and even image part
For testing the performance of each proposed CBIR
method, 55 queries (randomly selected 5 from each category) Figure 4 shows the performance comparison of proposed CBIR
are fired on the image database. The feature vector of query methods with the combination of original and even image parts. It is
image and database image are matched using the Euclidian observed that the proposed CBIR methods give better performance
distance. The average precision and recall values are found for with the combination of original and even image part than the
all the proposed CBIR methods. The intersection of precision original alone both in RGB and HSV color space. However an
exceptional behaviour has been observed in case of „4-pattern‟
and recall values gives the crossover point. The crossover
texture in HSV color space where original image outperforms the
point of precision and recall is computed for all the proposed combination of original and even image part.
CBIR methods. The one with higher value of crossover point
indicates better performance.

Figure 5. Performance comparison of the ‟16-pattern‟ texture based image
retrieval using tiled bitmaps with the combination of original and even image
Figure 3. Performance comparison of proposed CBIR methods in RGB and part
HSV color space

67 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
Figure 5 shows performance comparison of the ‟16-pattern‟ texture 2009, K.J.Somaiya College of Engineering, Vidyavihar, Mumbai-77.
based image retrieval using tiled bitmaps with the combination of [9] Minh N. Do, Martin Vetterli, “Wavelet-Based Texture Retrieval Using
original and even image part. It is observed that in case of HSV color Generalized Gaussian Density and Kullback-Leibler Distance”, IEEE
space the proposed CBIR methods do not give better performance Transactions On Image Processing, Volume 11, Number 2, pp.146-158,
with tiled bitmaps. The difference in the crossover points of „1Tile‟ February 2002.
and „4Tile‟ for the combination of original and even image part in [10] B.G.Prasad, K.K. Biswas, and S. K. Gupta, “Region –based image
HSV color space is negligible. Moreover the crossover point of retrieval using integrated color, shape, and location index”,
International Journal on Computer Vision and Image Understanding
original image in HSV color space with „1Tile‟ is higher than that
Special Issue: Colour for Image Indexing and Retrieval, Volume 94,
with „4Tile‟ bitmap. Issues 1-3, April-June 2004, pp.193-233.
VIII. CONCLUSION [11] Dr. H.B.Kekre, Sudeep D. Thepade, “Creating the Color Panoramic
View using Medley of Grayscale and Color Partial Images ”, WASET
As compared to the texture pattern based image retrieval using International Journal of Electrical, Computer and System Engineering
Walsh-Hadamard transform in RGB color space [30], the (IJECSE), Volume 2, No. 3, Summer 2008. Available online at
www.waset.org/ijecse/v2/v2-3-26.pdf.
performance of image retrieval can be ameliorated using the
[12] Stian Edvardsen, “Classification of Images using color, CBIR Distance
HSV color space. Moreover, it is observed that the Measures and Genetic Programming”, Ph.D. Thesis, Master of science
performance of proposed CBIR method improves with in Informatics, Norwegian university of science and Technology,
increasing number of texture patterns up to a certain level. The Department of computer and Information science, June 2006.
combination of original image with even image part gives [13] Dr. H.B.Kekre, Tanuja Sarode, Sudeep D. Thepade, “DCT Applied to
better performance than the original image alone. The Row Mean and Column Vectors in Fingerprint Identification”, In
Proceedings of International Conference on Computer Networks and
proposed CBIR methods in HSV color space do not give better Security (ICCNS), 27-28 Sept. 2008, VIT, Pune.
results with tiled bitmaps. Among the various texture patterns [14] Zhibin Pan, Kotani K., Ohmi T., “Enhanced fast encoding method for
used for content based image retrieval, 16 Walsh-Hadamard vector quantization by finding an optimally-ordered Walsh transform
texture patterns (16-pattern) in HSV color space give the best kernel”, ICIP 2005, IEEE International Conference, Volume 1, pp I -
result with the combination of original image and even image 573-6, Sept. 2005.
part, as indicated by the highest average precision-recall [15] Dr. H.B.Kekre, Sudeep D. Thepade, “Improving „Color to Gray and
Back‟ using Kekre‟s LUV Color Space”, IEEE International Advanced
crossover point value.
Computing Conference 2009 (IACC‟09), Thapar University, Patiala,
INDIA, 6-7 March 2009. Is uploaded at online at IEEE Xplore.
IX. REFERENCES
[16] Dr. H.B.Kekre, Sudeep D. Thepade, “Image Blending in Vista Creation
[1] Dr. H.B.Kekre, Sudeep D. Thepade, Varun K. Banura, “Augmentation using Kekre's LUV Color Space”, SPIT-IEEE Colloquium and
of Colour Averaging Based Image Retrieval Techniques using Even International Conference, Sardar Patel Institute of Technology, Andheri,
part of Images and Amalgamation of feature vectors”, International Mumbai, 04-05 Feb 2008.
Journal of Engineering Science and Technology (IJEST), Volume 2,
[17] Dr. H.B.Kekre, Sudeep D. Thepade, “Color Traits Transfer to
Issue 10, (ISSN: 0975-5462) Available online at http://www.ijest.info
Grayscale Images”, In Proc.of IEEE First International Conference on
[2] Dr. H.B.Kekre, Sudeep D. Thepade, Varun K. Banura, “Amelioration Emerging Trends in Engg. & Technology, (ICETET-08), G.H.Raisoni
of Colour Averaging Based Image Retrieval Techniques using Even COE, Nagpur, INDIA. Uploaded on online IEEE Xplore.
and Odd parts of Images”, International Journal of Engineering Science
[18] http://wang.ist.psu.edu/docs/related/Image.orig (Last referred on 23
and Technology (IJEST), Volume 2, Issue 9, (ISSN: 0975-5462)
Sept 2008)
Available online at http://www.ijest.info.
[19] Dr. H.B.Kekre, Sudeep D. Thepade, “Using YUV Color Space to Hoist
[3] Dr. H.B.Kekre, Sudeep D. Thepade, Akshay Maloo, “Query by Image
the Performance of Block Truncation Coding for Image Retrieval”,
Content Using Colour Averaging Techniques”, International Journal of
IEEE International Advanced Computing Conference 2009 (IACC‟09),
Engineering Science and Technology (IJEST), Volume 2, Issue 6,
Thapar University, Patiala, INDIA, 6-7 March 2009.
2010.pp.1612-1622 (ISSN: 0975-5462) Available online at
http://www.ijest.info. [20] Dr. H.B.Kekre, Sudeep D. Thepade, Archana Athawale, Anant Shah,
Prathmesh Verlekar, Suraj Shirke,“Energy Compaction and Image
[4] Dr. H.B.Kekre, Sudeep D. Thepade, “Boosting Block Truncation
Splitting for Image Retrieval using Kekre Transform over Row and
Coding using Kekre‟s LUV Color Space for Image Retrieval”, WASET
Column Feature Vectors”, International Journal of Computer Science
International Journal of Electrical, Computer and System Engineering
and Network Security (IJCSNS),Volume:10, Number 1, January 2010,
(IJECSE), Volume 2, Number 3, pp. 172-180, Summer 2008. Available
(ISSN: 1738-7906) Available at www.IJCSNS.org.
online at http://www.waset.org/ijecse/v2/v2-3-23.pdf
[21] Dr. H.B.Kekre, Sudeep D. Thepade, Archana Athawale, Anant Shah,
[5] Dr. H.B.Kekre, Sudeep D. Thepade, “Image Retrieval using
Prathmesh Verlekar, Suraj Shirke, “Walsh Transform over Row Mean
Augmented Block Truncation Coding Techniques”, ACM International
and Column Mean using Image Fragmentation and Energy Compaction
Conference on Advances in Computing, Communication and Control
for Image Retrieval”, International Journal on Computer Science and
(ICAC3-2009), pp. 384-390, 23-24 Jan 2009, Fr. Conceicao Rodrigous
Engineering (IJCSE),Volume 2S, Issue1, January 2010, (ISSN: 0975–
College of Engg., Mumbai. Is uploaded on online ACM portal.
3397). Available online at www.enggjournals.com/ijcse.
[6] Dr. H.B.Kekre, Sudeep D. Thepade, “Scaling Invariant Fusion
[22] Dr. H.B.Kekre, Sudeep D. Thepade, “Image Retrieval using Color-
of Image Pieces in Panorama Making and Novel Image
Texture Features Extracted from Walshlet Pyramid”, ICGST
Blending Technique”, International Journal on Imaging (IJI),
International Journal on Graphics, Vision and Image Processing
www.ceser.res.in/iji.html, Volume 1, No. A08, pp. 31-46, Autumn
(GVIP), Volume 10, Issue I, Feb.2010, pp.9-18, Available online
2008.
www.icgst.com/gvip/Volume10/Issue1/P1150938876.html
[7] Hirata K. and Kato T. “Query by visual example – content-based image
[23] Dr. H.B.Kekre, Sudeep D. Thepade, “Color Based Image Retrieval
retrieval”, In Proc. of Third International Conference on Extending
using Amendment Block Truncation Coding with YCbCr Color Space”,
Database Technology, EDBT‟92, 1992, pp 56-71
International Journal on Imaging (IJI), Volume 2, Number A09,
[8] Dr. H.B.Kekre, Sudeep D. Thepade, “Rendering Futuristic Image Autumn 2009, pp. 2-14. Available online at www.ceser.res.in/iji.html.
Retrieval System”, National Conference on Enhancements in Computer,
[24] Dr. H.B.Kekre, Tanuja Sarode, Sudeep D. Thepade, “Color-Texture
Communication and Information Technology, EC2IT-2009, 20-21 Mar

68 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
Feature based Image Retrieval using DCT applied on Kekre‟s Median
Codebook”, International Journal on Imaging (IJI), Volume 2, Number AUTHORS PROFILE
A09, Autumn 2009,pp. 55-65. Available online at
www.ceser.res.in/iji.html (ISSN: 0974-0627).
Dr. H. B. Kekre has received B.E. (Hons.) in Telecomm.
[25] Dr. H.B.Kekre, Sudeep D. Thepade, Akshay Maloo “Performance
Engineering. from Jabalpur University in 1958, M.Tech
Comparison for Face Recognition using PCA, DCT &WalshTransform
(Industrial Electronics) from IIT Bombay in 1960,
of Row Mean and Column Mean”, ICGST International Journal on
M.S.Engg. (Electrical Engg.) from University of Ottawa in
Graphics, Vision and Image Processing (GVIP), Volume 10, Issue II,
1965 and Ph.D. (System Identification) from IIT Bombay
Jun.2010, pp.9-18, Available online
in 1970 He has worked as Faculty of Electrical Engg. and
http://209.61.248.177/gvip/Volume10/Issue2/P1181012028.pdf..
then HOD Computer Science and Engg. at IIT Bombay. For
[26] Dr. H.B.Kekre, Sudeep D. Thepade, “Improving the Performance of 13 years he was working as a professor and head in the
Image Retrieval using Partial Coefficients of Transformed Image”, Department of Computer Engg. at Thadomal Shahani
International Journal of Information Retrieval, Serials Publications, Engineering. College, Mumbai. Now he is Senior Professor
Volume 2, Issue 1, 2009, pp. 72-79 (ISSN: 0974-6285) at MPSTME, SVKM‟s NMIMS University. He has guided
[27] Dr. H.B.Kekre, Sudeep D. Thepade, Archana Athawale, Anant Shah, 17 Ph.Ds, more than 100 M.E./M.Tech and several
Prathmesh Verlekar, Suraj Shirke, “Performance Evaluation of Image B.E./B.Tech projects. His areas of interest are Digital Signal
Retrieval using Energy Compaction and Image Tiling over DCT Row processing, Image Processing and Computer Networking. He
Mean and DCT Column Mean”, Springer-International Conference on has more than 320 papers in National / International
Contours of Computing Technology (Thinkquest-2010), Babasaheb Conferences and Journals to his credit. He was Senior
Gawde Institute of Technology, Mumbai, 13-14 March 2010, The paper Member of IEEE. Presently He is Fellow of IETE and Life
will be uploaded on online Springerlink. Member of ISTE Recently ten students working under his
[28] Dr. H.B.Kekre, Tanuja K. Sarode, Sudeep D. Thepade, Vaishali guidance have received best paper awards and two have been
Suryavanshi,“Improved Texture Feature Based Image Retrieval using conferred Ph.D. degree of SVKM‟s NMIMS University.
Kekre‟s Fast Codebook Generation Algorithm”, Springer-International Currently 10 research scholars are pursuing Ph.D. program
Conference on Contours of Computing Technology (Thinkquest-2010), under his guidance.
Babasaheb Gawde Institute of Technology, Mumbai, 13-14 March
2010, The paper will be uploaded on online Springerlink.
Sudeep D. Thepade has Received B.E.(Computer) degree
[29] Dr. H.B.Kekre, Tanuja K. Sarode, Sudeep D. Thepade, “Image from North Maharashtra University with Distinction in 2003.
Retrieval by Kekre‟s Transform Applied on Each Row of Walsh M.E. in Computer Engineering from University of Mumbai
Transformed VQ Codebook”, (Invited), ACM-International Conference in 2008 with Distinction, currently pursuing Ph.D. from
and Workshop on Emerging Trends in Technology (ICWET SVKM‟s NMIMS, Mumbai. He has about than 07 years of
2010),Thakur College of Engg. And Tech., Mumbai, 26-27 Feb 2010, experience in teaching and industry. He was Lecturer in
The paper is invited at ICWET 2010. Also will be uploaded on online Dept. of Information Technology at Thadomal Shahani
ACM Portal. Engineering College, Bandra(w), Mumbai for nearly 04
[30] Dr. H. B. Kekre, Sudeep D. Thepade, Varun K. Banura, “Image years. Currently working as Associate Professor in Computer
Retrieval using Texture Patterns generated from Walsh-Hadamard Engineering at Mukesh Patel School of Technology
Transform Matrix and Image Bitmaps”, Springer International Management and Engineering, SVKM‟s NMIMS University,
Conference on Technology Systems and Management (ICTSM 2011), Vile Parle(w), Mumbai, INDIA. He is member of
MPSTME and DJSCOE, Mumbai, 25-27 Feb 2011. The paper will be International Association of Engineers (IAENG) and
uploaded online on Springerlink. International Association of Computer Science and
[31] Dr. H. B. Kekre, Sudeep D. Thepade, Varun K. Banura, “Query by Information Technology (IACSIT), Singapore. He has been
Image Texture Pattern content using Haar Transform Matrix and Image on International Advisory Board of many International
Bitmaps”, Invited at ACM International Conference and Workshop on Conferences. He is Reviewer for many reputed International
Emerging Trends in Technology (ICWET 2011), TCET, Mumbai, 25- Journals. His areas of interest are Image Processing and
26 Feb 2011. The paper will be uploaded online on ACM portal. Computer Networks. He has more than 100 research papers
[32] Dr. H. B. Kekre, Sudeep D. Thepade, Varun K. Banura, “Image in National/International Conferences/Journals to his credit
Retrieval using Shape Texture Patterns generated from Walsh- with a Best Paper Award at International Conference
Hadamard Transform and Gradient Image Bitmaps”, International SSPCCIN-2008, Second Best Paper Award at ThinkQuest-
Journal of Computer Science and Information Security (IJCSIS), 2009 National Level paper presentation competition for
Volume 8, Number 9, 2010.pp.76-82 (ISSN: 1947-5500), Available faculty, Best Paper Award at Springer International
online at http://sites.google.com/site/ijcsis Conference ICCCT-2010 and second best project award at
Manshodhan 2010.
[33] Dr.H.B.Kekre, Sudeep D. Thepade, Shrikant Sanas, "Improving
Performance of multileveled BTC based CBIR using Sundry Color
Spaces”, CSC International Journal of Image Processing (IJIP), Volume Varun K. Banura is currently pursuing B.Tech. (CE) from
4, Issue 6, Computer Science Journals, CSC MPSTME, NMIMS University, Mumbai. His areas of
Press, www.cscjournals.org interest are Image Processing and Computer Networks. He
has 07 research papers in International Conferences/Journals
to his credit.

69 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Analysis and Comparison of Medical Image Fusion
Techniques: Wavelet based Transform and
Contourlet based Transform
C G Ravichandran,RVS College of Engg. & Tech, Dindigul, e-mail: cg_ravi@yahoo.com
R. Rubesh Selvakumar, Research Scholar, Anna University of
Technology,Tricirappalli, e-mail: gopikarubesh2009@rediffmail.com
S. Goutham, Surya Engineering College,Erode e-mail: gouthamsanjay00@gmail.com

Abstract various systems differ from each other. The CT and MRI are the most
commonly used image fusion because the CT image very clearly portrays
Medical Image Fusion provides additional information the human body bone tissues. The MRI image brings out the signals of
for diagnosis when a registered and overlaid multiple patient images. bone tissues and calcific point is weaker but the resolution of soft tissues
Multiple images from the same imaging modality or multiple is much better than CT. Both CT and MRI images are fused and they can
modalities can be used to create a fused image. The fused image is of completer the merits.[1].
immense help and provide more information to the doctor so as to
diagnose the diseases. the CT (Computer Tomography) offers less Many new algorithms of Medical Image Fusions have been developed in
detailed information for soft tissues and good information on bony the formation of two approaches since the 80’s of 20th century namely
structures. But the MRI (Magnetic Resonance Imaging) provides a Spatial Domain and Transform Domain. Averaging method, Brovey
more detailed information on the soft tissues but less detailed method, Principle Component Analysis(PCA) and intensity-hue-saturation
information on bone structures and the contrast resolution of soft (IHS) are some of the Spatial oriented Techniques. All the algorithms in
tissues is far better in MRI than in CT. In this paper, Wavelet based these techniques have the drawback of spatial distortion in the fused
Transform like DWT (Discrete Wavelet Transform) and CWT image. When we move on to further processing such as classification
(Complex Wavelet Transform) is analyzed theoretically and it is problemical distortion, the spatial distortion becomes a negative factor.[2]
compared with Contourlet based Transform like CCT (Complex But the multi-resolution tool which was initiated by burt and adelson in
Contourlet Transform) and NSCT (No-Subsampled Contourlet 1983 can overcome these drawbacks and it is known as Lapcian
Transform). The experimental results showing the evaluation Pyramid.[3] In continuous Function in all types of Pyramid based methods
measures like IE (Information Entrophy) , AG (Average Gradient) is in-proper and the decomposes process is very poor. The Wavelet
and SD (Standard Deviation). Transform (WT) as multi-resolution established by Mallrt and Meyar for
continuos function process good frequency division characteristics in the
Key Terms: CT (Computer Tomography), MRI (Magnetic Resonance transform region and this has rewulted in its wide use in the medical
Imaging), DWT (Discrete Wavelet Transform), CWT (Complex image fusion.[4]
Wavelet Transform), CCT (Complex Contourlet Transform), NSCT
(No-Subsampled Contourlet Transform). Though Wavelet based Transform solves the problems of low contrast and
blocking effects in space domain and avoid artifacts, they fail to achieve
I. INTRODUCTION good performance. All types of wavelet based transform provide
The process of combining relevant information from two or more images in-sufficient information like curve shape and edge representation
into a single image is known as Image Fusion. The fused image contains and also occurs the problems of shift in-variance and lack of
more informations than any one of the input images. Ariel and Satellite
directional selectivity. Then the need for a new approach arose
Imaging, Robot Vision, Remote Sensing and Medical Imaging are some
Muh D. Do and Vetterlin in 2002 proposed a new approach :
of the available image fusions. Nowadays, Medical Image Fusion
occupies a position of considerable importance in the field of medical contourlet Transform” known as multi-geometric Analysis.[5]. It
image clinical diagnosis. The Medical Image Fusion objectives is to happens to be a some powerful and useful tool for analyzing the
obtain a high resolution image replete with many details for the sake of signals consisting of lines, curves and edges than wavelet
diagnosis. transform. Its application to image fusion can provide sufficient
information to the doctor for diagnosis purposes. Several methods of
Several of various medical modalities images information that are contourlet transform such as Discrete Contourlet Transform(DCT),
comprehended together to form one image in order to express its Complex Contourlet Transfor(CCT) etc. These types of algorithms
information comprises Medical Image Fusion. As a result, the doctor is consists of some problesm such as lack of shift in-variance and
provided with a more effective diagnosis information by this image. X- directional selectivity. To overcome these problem prof. Cunha and Shun
Ray, Ultra Sound(US), ( PET) Scans (Positron Emission Tomography) was proposed the Non- Subsampled Contourlet Transform (NSCT).
and (SPECT) Scans (Single Photon Emission Computed Tomography) ,
CT(Computer Tomography) and MRI (Magnetic Resonance Imaging( are II WAVELET TRANSFORM
the various modality Images, that are in use for clinical analysis and In order to identify the vital details in the image the registered image is
treatment. The single modality image cannot often give complete and applied by the transform in general concept of transform domain fusion. .
accurate information for this doctor since the formation principles of First any fusion rule is applied to the transform co-efficient and then

1
70 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

obtained to the fusion decision map, then applied to the decision map by high-low(HL), los-high(LH) and low-low(LL) image subbands.[9] By
the transform domain. Finally, the resulted image will have more details recursively applying the same scheme to the low-low subband a multi-
of both the source images.[6]. resolution decomposes process can be achieved.(Fig.3)

Fusion Fusion Fused
decisio Transfor Image LH1 HH1
m co-
n map
efficent

Registered Transform
Image co-efficients
Fig. 1. Block diagram of Transform domain image fusion LH2 HH2
The main concept and theory of wavelet based multi-resolution is
delivered from mallrt. It can detect local features in a signal process and it
si used to decompose process. Besides texture analysis, data compression
and feature detection, it can also be used for image fusion. Even before LH3 HH3 HL1
wavelet based transform, the pyramid base transform was introduced by
Burt and Adelson in 1983 and then it was improvised by Toet [7]. But in Fig.3. Labelled subbands
these type of techniques are not supported to continuous function. So the
wavelet technique is applied to the image fusion. Fusion Rules: there are three fusion rules that were developed and
implemented using DWT based image fusion.
In common , all types of transform domain fusion techniques , the
transformed image are combined in the transform domain using any of the 1. maximum selection (MS) scheme: This simple scheme just
picks the coefficient in each subband with the largest
fusion rule, it is denoted by of the registered input images I1 (x,y) and I2
magnitude;
(x,y) together with the fusion rule φ then the inverse wavelet transform W- 2. weighted average (WA) scheme: This scheme developed by
1
is applied and the fused image I (x,y) is reconstructed.[8] Burt and Kolczynski [10] uses a normalised correlation
between the two images’ subbands over a small local area. The
I(x,y) = w-1 (Φ(w(I1 (x,y)), w(I2 (x,y)))) resultant coefficient for reconstruction is calculated from this
measure via a weighted average of the two images’
……………… (1) coefficients.

3. window based verification (WBV) scheme: This scheme
developed by Li et al. creates a binary decision map to choose
I1 between each pair of coefficients using a majority filter.
w-1 In DWT fits most commonly into the decimated form(Mallats Dyadic
Filter Tree0[11]. It is only used for compression but other occus in other
signal analysis such as lack of shift in-variance. Sice the wavelet fitness
are separable, the small shift in the input signal can cause major variations
I2 in the energy distribution between DWT coefficients at different scales
and poor directional selectivity for diagonal features. So the new approach
was introduced to shift in-variance tha is known as Complex Wavelet.
However, the real valued wavelet transform suffers from shift variance
Registered wavelet fused wavelet Fused and lack of directional selectivity. Nikolov et al. [12] introduced the use of
image the dual tree complex wavelet transform (DT-CWT) for image fusion. The
Input Images coefficients coefficients Images DT-CWT is approximately shift invariant and has double the amount of
directional selectivity compared to a real valued wavelet transform. Shift
invariance is an important feature of a fusion transform as the magnitude
Fig.2. Fusion of the Wavelet Transform of two images of the coefficients of a shift variant transform may not properly reflect
their importance. The improved directional selectivity of the DT-CWT is
III DISCRETE WAVELET TRANSFORM also important in order to properly reflect the content of the images across
boundaries and other important directional features. In DT-CWT gives
The main idea of all multi-resolution schemes centers around the human
much better directional selectivity when multi-dimensional signal is
vital system being local contrast ie., edges or corners. Two registered filtered.[10].
images of the same scene each of the coefficients of each transform
possess significantly different magnitudes within the region Using any
one of the fusion rule to generate the combined coefficients map then the IV. DUAL-TREE COMPLEX WAVELET TRANSFORM
inverse DWT is applied to the combined coefficients map to produce the
fused image which is give the more information of than the input images. The motivation of suing the DT-CWT for image fusion is, reduced the
Using separate Filter and down-sampling in the horizontal and vertical shift in-variance and directional selectivity when compared with the
DWT. In DT-DWT comprising two trees of real filter a and b, which
directions produces four subbands at each scale. It denotes the horizontal
produce the real and imaginary parts of the complex coefficients and odd
frequency and then the vertical frequency. This produces high-high(HH), and even length bi-orthogonal linear phase filters.

2
71 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

where Igt is the cut-and-paste “ground truth” image, fd is the fused image
and N is the size of the image. Lower values of _ indicate greater
similarity between the images Igt and Ifd and therefore more successful
fusion in terms of quantitatively measurable similarity. Table 1. shows the
results for the various methods with several fusion rules are used.

S.No. Methods
1. DWT- MS Fusion Rule 8.2964
2. DWT-WA Fusion Rule 7.6551
3. DWT-WBV Fusion Rule 7.5271
4. DT-CWT- MS Fusion Rule 7.2284
5. DT-CWT- WA Fusion Rule 7.2043
Figure 4: Dual tree of real filters for the CWT, giving real and 6. DT-CWT-WBV Fusion Rule 6.9540
imaginary parts of complex coefficients Table 1. Quantitative results for various fusion methods

Unfortunally , the odd/even length filter approach suffers from certain There arised some problems in all wavelets based transforms such as the
problems [13] such as a. sub-sampling structures is not very symmetrical accuracy of edgeand curve localization in the wavelet transform is low
b. two trees have slightly different property response and c. the filter sets and it has paved the way for an alternative approach which has a high
must be bi-orthogonal. To overcome this problem the same author N. G. accuracy of curve localization such as the contourlet transform The first
Kingbury slightly modified the existing algorithm to a developed one MGA tool was proposed by muh D. Do and Martin Vetterli in 2002
which is Q-Shift DT-CWT for image fusion.[13,14]
[15].The main idea of this method is to construct a multiresolution and
multidirection (MRMD) decomposed representation of image and also it
is more powerful tool for the analysis of the signals consists of lines,
curves and edges than wavelet transform. It applied to image fusion, can
provide more information to the doctor for diagnosis purposes. [16].
Thereare several methods in MGA such as discrete contourlet transform ,
complex contourlet transform. These transform ae lack of shift
invariance,to overcome this problem Prof. cunha and zhna was proposed
non-subsampled contourlet transform(NSCT). It is a fully shift invariant ,
multiscale and multidirectional transformation. [17] In this paper, we
present the comparative and evaluative study of these three contourlet
transform.
(a) MRI Image (b) CT Image
V. DISCRETE CONTOURLET TRANSFORM
The Contourlet Transform is a new geometrical image transform , which
represents image formation containing contours and textures. The property
of contourlet is directionality and anisotropy, first contourlet transform
was introduced by Do and Vetterli. A morepowerful MSMD framework is
Contourlet Transform, that consists of Laplacian Pyramid(LP) and
Directional Filter Bank(DFB). The LP decomposes on image into
subbands and DFB is an analysis of each digital image.

(c ) Fused Image –DWT (d) Fused Image- CT-CWT

Fig. 5(a,b,c) Input and Output Images Fig 6. A Flow graph of the Contourlet Transform

For the Contourlet Transform , first a standard multiscale decomposition
QUANTITATIVE COMPARISONS into different bands is computed, where the lowpass channel is subbandled
while highpass is not. Then a directional decomposition with a DFB is
Under the circumstances, comparisons of quantitative quality tends to be applied to each highpass channel. The DFB is a critically sampled filter
misleading or meaningless. But a few authors have tried to formulate such bank that can decompose on images of directions. So, one can decompose
measures for applications with clear meaning. This is produced using a each scale into any arbitrary power of two’s number of directions. Before
simple cut-and-paste technique, physically taking the “in focus” areas contourlet decomposition,registration of the input images must be done.
from each image and combining them. The quantitative measure used to
compare the cut-and-paste image to each fused image was taken from There are three basic steps for the proposed contourlet based image
fusion are, 1. The input Images A and B to be decomposed using
decompose images into lowpass coefficients and bandpass coefficients . 2.
To combine the transform coefficients based on two types of way, one is
fusion rules based on pixel and fusion rule based on Region. And 3.
Finally, using contourlet transform to construct the fused images.

3
72 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Fig. 7. Framework of Contourlet Transform
Fig. 9. Procedure of the Complex Contourlet Transform (CCT)

CCT comprises three steps for the fusion of images. Subband and
directional decomposes are used in this proposed transform. Subsequently
followed by applying certain fusion rules and the transform coefficients .
finally, the complete coefficients are used with maximum magnitudes.
The ending of complete fusion scheme applies inverse complex contourlet
transform. The salient features of an image is edges and curve
boundaries.

Fig. 8. The Fusion Framework based on Contourlet Transform

The Contourlet Transform satisfy anisotropy principle and can capture
intrinsic geometric structure information of images. It achieve better
expression than discrete wavelet transform, especially meant for edges
and contours. It can be utilized for extracting the geometric information Fig.10. the proposed fusion scheme using CCT
of images very well, and it is usefulto many image processing tasks. Small
features challenge the contourlet that represents the long edges. The
contourlet transform lacks shift in-variance due to the down- sampling and Based on quantitative Measurements such as, Mean Gradient
up-sampling. Hence the read for another approach for the purpose of (MG), Average Value(AV) and Correlation coefficient (CC), I
problem solving has risen [18] [19]. present the previous experimental results using the PCA based fusion,
SIDWT based fusion, DT-CWT based fusion and finally CCT Transform
VI. COMPLEX CONTOURLET TRANSFORM based fusion scheme. [22].

Complex Contourlet Transform incorporates the DT-CWT because Method MG CC AV
after investigation, the DT-CWT tasks advantages of approximate shift in- PCA 724.6968 0.8446 78.6876
variance and directional selectivity for image fusion. But , the DT-CWT SIDWT 1041 0.8803 79.0012
can only handle limited direction informations. So, the researcher Dipeng DT-CWT 1143.9 0.8952 80.1860
Chen and Qi Li proposed one method called as Complex Contourlet CCT 1147.9 0.9166 80.2766
Transform(CCT). It provides simultaneous better directional sensitivity Taable.2. Quantitative results for varius fusion methods
and shift invariance. CCT consists of two steps, first one is, A CT-CWT The CCT based fusion methods has higher spectral quality compared
decomposes, this is the contrast to the critically used in sampled DWT with the other methods, in terms of the higher values of correlation
[20] and Lappacian Pyramid [21]. After applying DT-CWT decomposes coefficients and mean gradient.
then only apply the CCT. It allows for of each scale arbitrary directions
and also approximate shift invariance. VII. NON-SUBSAMPLED CONTOURLET
TRANSFORM(NSCT)

Previous contourlet transformmethods satisfy anisotropy
principles by capturing geometric structures information of
images. The image expressed is better especially for edges and
contours. However during the down sampling and up-sampling
produce lack of shift in-variance and artfacts. So, Cunha and

4
73 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Zhon proposed another method which is called as Non-
Subsampled Contourlet Transform(NSCT). [23]. NSCT means,
combination of Non-Subsampled Pyramid provides multi-scale
decomposes and Non-Subsampled Directional Filter Banks provides
directional decomposes. [24].

Fig11.. sampling filters by a quincunx Q C.Contourlet Transform D. Complex Contourlet Transform

Fig12.. iteratal non-subsampled directional filter
E. Non-Subsampled Contourlet Transform
In this method it can be repeatedly iterated on lowpass subband outputs Fig. 14. Input and Output Images
of non- subsampled pyramids. First, non-subsampled pyramid, the input is
split into a lowpass subband and a highpass subband, then only the DFB VIII. QUANTITATIVE ANALYSIS
decomposes the highpass subband into several directional subband.
Finally, it is iterated repedeatelly, on the lowpass subband. There are three evaluation measures are used in this paper, as follows:

1) Information entropy (IE): The IE of the image is an important index
to measure the abound degree of the image information. The larger the IE
is, the more information the image carries. The IE of the image is
definition as

(1)

Where the h(i) is ratio of the number of the pixels with gray value equal to
Fig 13. .a) muliscale decomposes/ directional decomposes i over the total number of the pixels.
fig. b)idealized frequency portioning
2) Average Gradient (AG): The definition is denoted as
The NSCT provides not only muliresolution analysis but also
geometrical and directional representation. The NSCT is shift invariant
and more powerful tool for image fusion and it provides better result
compared with the CT and CCT..

Images images

(2)
The average gradient reflects the contrast degree in
detail and texture change. The larger of the AG is, the more clear the
image is.

4) Root Mean Square Error (RMSE) : Suppose r is the source image
(standard reference image) and f is the fused image; the root mean square
error is defined as follows:

Output Images
(3)

The RMSE is used to measure the difference between the source image
and the fused image; the smaller the value of RMSE and the smaller the
difference, the better the fusion performance.

5
74 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

[7]. Mnh N. Do. Martic Vetterli, “ The Contourlet Transform: An
Transform IE AG RMSE Efficient Directional Multi-resolution Image Representation”, IEEE
Transaction on image processing, vol.14, no.12, pp. 2001-2106, 2005.
Contourlet 5.616 5.270 15.56
Transform [8]. Paul Hill, Nishan Canagarajah and Dave Bull, Dept. of Electrical and
Complex Electronic Engine ering, The University of Bristol, Bristol, BS5 1UB,
contourlet 5.717 5.401 17.23 UK,” Image Fusion using ComplexWavelets”,
Transform
Nonsubsampled [9]. [kingbury, 1998] kingbury N.G. (1998). The dual-tree complex
contourlet 6.170 5.462 15.43 wavelet transform: a new technique for shift in-variant and directional
Transform filters, IEEE digital signal processing workshop (86).
Table 3 Comparison of fusion Results
[10]. S.G. Mallar,” A Theory for multi-resolution Signal Decomposition:
From the table above we can see that the Nonsubsampled the Wavelet representation”, IEEE Transaction on PAMI, 11)7), pp. 674-
contourlet transform have the higher value of Information entropy, 693, 1989.
Average gradient and lower value of root mean square error for the fused
image. It is shown that the NSCT provide the better fusion result. [11].S. Nikolov, P.R. Hill, D.R. Bull, C.N. Canagarajah, “Wavelets for
image fusion,” in Wavelets in Signal and Image Analysis, from Theory to
XI. EXPERIMENT RESULT Practice, A. Petrosian and F. Meyer, Eds. Kluwer Academic Publishers,
2001
We use parietal image of CT and MR for experiment analysis. Fig 8(a)
show the CT image and fig 8(b) shows the MR image .From the figure, [12]. N.Kingbury,” A Dual Tree Complex Wavlet Transform with
We can see that the CT image does not shows the soft tissues clearly and improved orthogonal and symmetry properties”, ICIP, vol2, 2000. Pp.
In MR image, the soft tissues are clear but it does not show the coronal 375-378.
bones clearly. The experiment compared the contourlet transform, with
complex contourlet transform and nonsubsampled contourlet transform. [13]. N.Kingbury, “ Design of Q-Shift Complex Wavelet for image
The Fig.8(c,d,e) is the result of fusion method using the above three Processing using frequency domain energy minimization”, International
contourlet transforms. Conference on Image Processing, 2003(1): 1013-1016.
X. CONCLUSION
[14]. Minh N. Do, Martin Vetterli,” The Contourlet Transform: An
This paper presents the comparison of Wavelet Based Transform efficient Directional Multiresolution Image Representation,” IEEE
contourlet transform in terms of various performance measures,like Transaction on Image Processing”, Vol.14. No. 12 pp. 2001-2106, 2005.
Quantitative Measure its used for Wavelet Transform and Information
Entropy (IE), Average Gradient (AG) and Root Mean Square [15]. YangChai, Yante, Chaolong Ying, “ CT and MRI Image Fusion
Error(RMSE).. Non-Subsampled Contourlet Transform(NSCT ) its used based on Contourlet using a Novel Rule”, IEEE Transaction on Image
for Contourlet Transform, provides very good results for pixel level Processing.
fusion due to its improved directionality , better geometric representation
and good directional selectivity . Hence using the NSCT, one can have [16]. C.S. Pattichis, M.S. Pattrchis, “ Adaptive Normal Network Imaging
the fused image with better direction, high geometric representation and in Medical System,” Proceedings of the 35th Asilomar Conference on
better visual quality. This will help the physicians to diagnose diseases in Signal, System and Computer”, Vol.1, pp. 313-317, 2001.
a better way.
[17]. [10]. R. Eslami and H. Radha, “ New Images Transform using
Hybrid wavelets and Directional Filter Banks: Analysis and Design,”
XI. REFERENCES Proc. IEEE Int. Conf. Image Processing, 2005, pp. 733-736
[1]. Ge Wen, Gao Li-Qun, Ge-Wen, “ A CT/MRI image fusion algorithms
combined non-separable wavelet transform and regional priority”, 2008 [18]. V. Vhappelier, C. Guillemot, and S. Marinkovic, “ Image coding
international symposium on computer science and computational with iterated contourlet and wavelet transform”, Proc. IEEE Int. Conf.
technology. Image Processing, 2004, pp. 3157-3160.

[2]. T.M. Tu, W.C. Cheng, C.P. Cheng, P.S. Hirang, “ Best Tradeoff for [19]. P.Piello, “ A general Framework for multi-resolution Image Fusion:
high resolution image fusion to preserve spatial details and minimize color from Pixels to Regions,” PNA-RCW11, ISSN 1386-3711, 2002
distortion”, IEEE Geoscience and Remote Sensing letters, 4 (2) 2007,
302-306. [20]. B. Jeon and D.A> Landrebe,” Decision Fusion Approach for multi-
temporal classification,” IEEE Transaction on Geoscience and Remote
[3]. G. Piellar, “ A General Framework for multi-resolution image fusion: Sensing, Vol.37. No.7. pp. 1227-1233, 1999.
from pixel to region information fusion”, Vol.4. pp.259-280, 2003.
[21]. Dipeng Chen and Qu Li , “ The Use of Complex Countourler
[4]. S.G. Mallert, “ Multi-frequency channel decomposition of image and Transform on Fusion Scheme”, World Academy of Science, Engineering
wavelet models”, IEEE Transaction on Acoustic , Speech and signal and Technology 12, 2005.
processing, vol.37, no.12, p.p. 2091-2110, 1989.
[22]. O. Rockernger, “ Image Sequence Fusion using a Shift invariant
[5]. Mnh N. Do. Martic Vetterli, “ The Contourlet Transform: An wavelet transform,” in Proc. Int. Conf. Image Processing, Washington,
Efficient Directional Multi-resolution Image Representation”, IEEE 1997, Vol.3, p.p. 288-291
Transaction on image processing, vol.14, no.12, pp. 2001-2106, 2005.
[23]. Jlanpring Zhou, Aruther C. Cunho and Minh N. Do, “ Non-
[6]. Y. kirankumar, “ Comparison of Fusion Techniqueapplied to Subsampled Contourlet Transform: Construction and Application in
practical images: Discrete Curvlet Transform using Wrapping Technique Enhancement”, IEEE Transaction on Image Processing
and Wavelet Transform”, Journal of theoretical and applied Information
Technology.

[7]. Yang Yang, Dang Sin Park, Shuying, Zhijun Eang,’ Wavelet based
approach for Fusion of CT and MRI”,.

6
75 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011

Performance Comparison of Texture Pattern Based
Image Retrieval Methods using Walsh, Haar and
Kekre Transforms with Assorted Thresholding
Methods
Dr. H.B.Kekre1, Sudeep D. Thepade2, Varun K. Banura3
1
Senior Professor, 2Ph.D.Research Scholar & Associate Professor, 3B.Tech (CE) Student
Computer Engineering Department, MPSTME,
SVKM‟s NMIMS (Deemed-to-be University), Mumbai, India
1
hbkekre@yahoo.com, 2sudeepthepade@gmail.com,3varunkbanura@gmail.com

Abstract— Novel texture pattern based image retrieval could be listed as art galleries [15,17], museums, archaeology
techniques using image maps and non-sinusoidal orthogonal [6], architecture design [11,16], geographic information
image transforms is the theme of the work presented in this systems [8], weather forecast [8,25], medical imaging [8,21],
section. Different texture patterns namely ‘4-pattern’, ‘16- trademark databases [24,26], criminal investigations [27,28],
pattern’, ‘64-pattern’ are generated using Haar transform image search on the Internet [12,22,23]. The paper attempts to
matrix, Walsh transform matrix and Kekre transform matrix. provide better and faster image retrieval techniques.
The generated texture patterns are then compared with the
image maps (binary image maps for Walsh patterns and Ternary A. Content Based Image Retrieval
image maps for Haar patterns & Kekre patterns) of an image to
generate the feature vector based on structural matching (as the For the first time Kato et.al. [7] described the experiments
matching number of ones, minus ones per Walsh texture pattern of automatic retrieval of images from a database by colour and
and matching number of ones, zeros, minus ones per Haar/Kekre shape feature using the terminology content based image
texture pattern). Further the image maps are created using four retrieval (CBIR). The typical CBIR system performs two major
thresholding methods as global, local, intermediate with 4 tiles tasks [19,20] as feature extraction (FE), where a set of features
(intermediate-4) and intermediate with 9 tiles (intermediate-9). called feature vector is generated to accurately represent the
Here total 36 variations of proposed novel image retrieval content of each image in the database and similarity
methods using texture patterns are considered with three image measurement (SM), where a distance between the query image
transforms (Walsh, Haar & Kekre), three variations in number and each image in the database using their feature vectors is
of texture patterns (4, 16 & 64) and four different ways to used to retrieve the top “closest” images [19,20,29].
generate image maps (with global, local, intermediate-4,
intermediate-9 thresholding methods). The proposed texture For feature extraction in CBIR there are mainly two
content based image retrieval (CBIR) techniques are tested on approaches [8] feature extraction in spatial domain and feature
the image database with help of 55 queries (randomly selected 5 extraction in transform domain. The feature extraction in
from each of 11 image categories) fired on image database. The spatial domain includes the CBIR techniques based on
performance comparison of texture pattern based CBIR methods histograms [8], BTC [4,5,19], VQ [24,28,29]. The transform
is done with help of precision-recall crossover points. domain methods are widely used in image compression, as they
give high energy compaction in transformed image [20,27]. So
Keywords- CBIR; Walsh, Haar & Kekre Transforms; it is obvious to use images in transformed domain for feature
Texture; Patterns; Image Maps extraction in CBIR [26]. But taking transform of image is time
consuming. Reducing the size of feature vector using pure
I. INTRODUCTION image pixel data in spatial domain and getting the improvement
Today the information technology experts are facing in performance of image retrieval is shown in [1,2,3]. But the
technical challenges to store/transmit and index/manage image problem of feature vector size still being dependent on image
data effectively to make easy access to the image collections of size persists in [1,2,3]. Here the query execution time is further
tremendous size being generated due to large numbers of reduced by decreasing the feature vector size further and
images generated from a variety of sources (digital camera, making it independent of image size. Many current CBIR
digital video, scanner, the internet etc.). The storage and systems use the Euclidean distance [4-6,11-17] on the extracted
transmission is taken care of by image compression [4,7,8]. feature set as a similarity measure. The Direct Euclidian
The image indexing is studied in the perspective of image Distance between image P and query image Q can be given as
database [5,9,10,13,14] as one of the promising and important equation 1, where Vpi and Vqi are the feature vectors of image
research area for researchers from disciplines like computer P and Query image Q respectively with size „n‟.
vision, image processing and database areas. The hunger of
superior and quicker image retrieval techniques is increasing
day by day. The significant applications for CBIR technology

76 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
n
ED   (Vpi  Vqi ) 2
(1)  1, if .B(i, j )  TB

i 1
BMb(i, j )   (8)
 1, if .B(i, j )  TB

II. GENERATION OF IMAGE MAPS
Image bitmaps are prepared using four different types of
thresholding considerations as global, local, intermediate-4
and intermediate-9. For global thresholding, the image maps B. Generation of Global Ternary Image Maps
of colour image are generated using three independent red (R), The ternary image maps are used for comparison with
green (G) and blue (B) components of image to calculate three texture patterns generated using Haar and Kekre transforms as
individual colour thresholds and one overall luminance the patterns contain three values one, zero and minus-one.
threshold [35]. Let X={R(i,j),G(i,j),B(i,j)} where i=1,2,….m Here first for each component (R, G, and B), the individual
and j=1,2,….,n; be an m×n color image in RGB space. Let the colour threshold intervals (lower-Tshl, and higher-Tshh) are
individual colour thresholds be TR, TG and TB which could calculated as shown in equations 9, 10 and 11.
be computed as per the equations given below as 2, 3 and 4.
Let the luminance threshold T be as given by equation 5.
Tshrl  TR  TR  T , Tshrh  TR  TR  T (9)
m n
1
TR   R(i, j)
m * n i 1 j 1
(2)
Tshgl  TG  TG  T , Tshgh  TG  TG  T (10)

1 m n
TG   G(i, j)
m * n i 1 j 1
(3) Tshbl  TB  TB  T , Tshbh  TB  TB  T (11)

Then the individual color plane global ternary image maps
1 m n
TB   B(i, j)
m * n i 1 j 1
(4)
are computed (TMr, TMg and TMb) as given in equations 12,
13 and 14. If a pixel value of respective color component is
greater than the respective higher threshold interval (Tshh),
the corresponding pixel position of the image map gets a value
TR  TG  TB „one‟; else if the pixel value is lesser than the respective lower
T (5)
threshold interval (Tshl), the corresponding pixel position of
3
the image map gets a value of „minus one‟; otherwise it gets a
value „zero‟.
A. Generation of Global Binary Image Maps
 1, if .R(i, j )  Tshrh
In binary image maps using global thresholding, the image 
TMr (i, j )   0, if .Tshrl  R(i, j )  Tshrh
is converted in to ones and minus ones only. Binary image  1, (12)
 if .R(i, j )  Tshrl
bitmaps are generated using the individual color component
threshold values (TR, TG, TB) as BMr, BMg and BMb. If a
pixel in each component (R, G, and B) is greater than or equal
to the respective threshold, the corresponding pixel position of  1, if .G(i, j )  Tshgh

the bitmap will have a value „one‟ otherwise it will have a TMg (i, j )   0, if .Tshgl  G(i, j )  Tshgh
value „minus one‟, as given by equations 6, 7 and 8. The  1, if .G(i, j )  Tshgl (13)

binary image maps are used for comparison with texture
patterns generated using Walsh transform.

 1, if .B(i, j )  Tshbh
 1, if .R(i, j )  TR 
 TMb(i, j )   0, if .Tshbl  B(i, j )  Tshbh
BMr(i, j )   (6)  1, (14)
 if .B(i, j )  Tshbl
 1, if .R(i, j )  TR


 1, if .G(i, j )  TG C. Generation of Global Ternary Image Maps

BMg(i, j )   (7) The binary or ternary image maps are generated using four
 1, if .G (i, j )  TG
 sundry methods of thresholding as global, intermediate-4,
intermediate-9 and local. For global thresholding based image

77 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
maps, the global threshold values are computed as average of (4-pattern). The 4x4 Walsh transform matrix is given in 2(d)
all pixel intensity values in the respective plane of considered and visualization of 16 Walsh transform patterns generated
colour image (as given by equations 2,3,4 and 5). In case of using it is shown in 2(e), where black color represent the
intermediate-4 thresholding, the image is divided into four values „1‟ in the pattern and values „-1‟ are represented by
equal non overlapping parts (as shown in (b) of figure 1) and white color. The obtained Walsh texture patterns then are
the image map of each of the part is generated using average resized as the size of image for which texture features have to
of pixels of only that part as threshold. In case of intermediate- be extracted.
9 thresholding based image maps, the image is divided into
nine non overlapping equal parts. For local thresholding, each
non distinct 2x2 pixel window is considered separately. Figure
1 shows the pixel group consideration for respective
thresholding methods, where the gray shading indicates the
group of pixel values to be considered for threshold
calculation in respective thresholding methods and number of
such pixel groups possible are given by black lines. (a) 2x2 Walsh (b) 2x2 Walsh (c) Generated 4
Matrix Matrix Walsh Texture
Patterns (4-pattern)

(a) Global (b) Intermediate-4

(d) 4x4 Walsh (e) Generated 16 Walsh Texture Patterns
Matrix (16-pattern)

Figure 2. Walsh Texture Pattern Generation
B. Haar Texture Pattern Generation
(c) Intermediate-9 (d) Local The 2x2, 4x4 and 8x8 Haar transform matrices are used to
Figure 1. Pixel group consideration for respective thresholding generate the 4, 16 and 64 Haar texture patterns respectively.
methods considered in image map generation for CBIR with The generation of four and sixteen Haar texture patterns [32]
texture patterns is shown in figure 3. 2x2 Haar transform matrix [9,30,31] is
shown as 3(a), each row of this matrix is considered one at a
III. TEXTURE PATTERN GENERATION time and is multiplied with all rows of the same matrix to
Using the non-sinusoidal transform matrices assorted generate Haar texture patterns as shown in 3(b). Figure 3(c)
texture patterns namely 4-pattern, 16-pattern and 64-pattern gives the visualization 4 Haar texture patterns (4-pattern). The
are generated. To generate N2 texture patterns (N2-pattern) 4x4 Haar transform matrix is given in 3(d) and 16 Haar
texture patterns, NxN transform matrix is considered and the transform patterns generated using it, are shown in 3(e), where
element wise multiplication of each row of the transform black colour represent the values „1‟ in the pattern, grey colour
matrix is taken with all possible rows of the same matrix represents values „0‟ and values „-1‟ are represented by white
(consideration of one row at a time gives one pattern). The colour. The obtained Haar texture patterns then are resized as
texture patterns obtained are orthogonal in nature. The the size of image for which texture features have to be
generation methods of Walsh transform, Haar transform and extracted. All the generated texture patterns are orthogonal to
Kekre transform patterns are elaborated respectively in each other.
sections A, B and C as given below.
A. Walsh Texture Pattern Generation
The 4, 16 and 64 Walsh texture patterns are generated using
Walsh transform matrices [21,22,26,36] of size 2x2, 4x4 and
8x8 respectively. The generated four and sixteen Walsh
texture patterns [34] are shown in figure 2, 2x2 Walsh
transform matrix is shown as 2(a), each row of this matrix is (a) 2x2 Haar (b) 2x2 Haar (c) Generated 4
considered one at a time and is multiplied with all rows of the Matrix Matrix Haar Texture
same matrix to generate Walsh texture patterns as shown in Patterns (4-pattern)
2(b). Figure 2(c) gives the envisioned 4 Walsh texture patterns

78 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
IV. CBIR USING TEXTURE PATTERNS
In all total thirty six variations of proposed CBIR method
are possible using the four methods of image maps (alias local,
global, intermediate-4 and intermediate-9), three image
transforms (Haar, Kekre and Walsh) and three different sets of
texture patterns (4-pattern, 16-pattern and 64-pattern). For
feature extraction in CBIR using texture patterns first the
image map is generated using appropriate thresholding method
(d) 4x4 Haar (e) Generated 16 Haar Texture Patterns
(local or global or intermediate-4 or intermediate-9). Then the
Matrix (16-pattern)
desired texture pattern set is generated (4-pattern or 16-pattern
or 64-pattern) using the corresponding image transform (Haar
Figure 3. Haar Texture Pattern Generation
or Kekre or Walsh).
C. Kekre Texture Pattern Generation
To generate feature vectors the binary image map of the
The 4, 16 and 64 Kekre texture patterns are generated using image is compared with each pattern of the generated Walsh
Kekre transform matrices [33,36] of size 2x2, 4x4 and 8x8 texture patterns to find matching number of ones & minus
respectively. Figure 4 gives generation of four and sixteen ones in case of CBIR with Walsh texture patterns. The feature
Kekre texture patterns. 2x2 Kekre transform matrix is shown vactor will have two values (number of matching „1‟ & „-1‟)
as 4 (a), each row of this matrix is considered one at a time per pattern per colour plane in Walsh texture patterns. The per
and is multiplied with all rows of the same matrix to generate image feature vector size for Walsh texture pattern based
Kekre texture patterns as shown in 4 (b) with all the negative CBIR is given by equation 15.
values are replaced by „-1‟. Figure 4 (c) gives visualization of
the 4 Kekre texture patterns (4-pattern). The 4x4 Kekre
transform matrix is given in 5.16 (d) and visualization of 16 Feature vector size=2*3*(no. of considered texture-pattern) (15)
Kekre transform patterns generated using it is shown in 5.16
(e), where black colour represent the values „1‟ in the pattern, In case of Haar or Kekre texture patterns based CBIR, he
grey colour represent the values „0‟ and values „-1‟ are ternary image map of the image is compared with each pattern
represented by white colour. The obtained Kekre texture of Haar or Kekre texture patterns to find three values per
patterns then are resized as the size of image for which texture colour plane per pattern as number of matching ones, zeros &
features have to be extracted. minus ones. The feature vector is formed using all these
number of matches (for „1‟, „0‟ and „-1‟). The size of the
feature vector of the image for Haar or Kekre texture patterns
based CBIR is given by equation 16. Table 1 shows the feature
vector size for 4, 16 and 64 texture patterns of respective
image transforms.

Feature vector size=3*3*(no. of considered texture-pattern) (16)
(a) 2x2 Kekre (b) 2x2 Kekre (c) Generated 4
Matrix Matrix Kekre Texture
Patterns (4-pattern) Table 1. Feature vector of image retrieval using texture paterns
16- 64-
CBIR Technique 4-Pattern
Pattern Pattern
Walsh Texture
8 32 128
Patterns
Haar Texture
12 48 192
Patterns
Kekre Texture
12 48 192
Patterns
(d) 4x4 Kekre (e) Generated 16 Kekre Texture Patterns
Matrix (16-pattern)
Using three assorted texture pattern set generated using
Figure 4. Kekre Texture Pattern Generation Walsh, Haar and Kekre transform matrices along with image
maps formed by different thresholds namely global,
intermediate-4, intermediate-9 and local, total 36 novel feature
vector generation methods have been tested resulting into six
new image retrieval techniques. The main advantage of
proposed CBIR methods is reduced time complexity for query
execution due to reduced size of feature vector resulting into

79 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
faster image retrieval with better performance. Also the pattern based CBIR methods with respective image map
feature vector size is independent of image size in proposed thresholding techniques are shown in figure 6. In CBIR using
CBIR methods. texture patterns intermediate-4 thresholding has given better
performance than other considered thresholding methods. For
V. IMPLEMENTATION each thresholding method 16 pattern has given better
The implementation of the discussed CBIR techniques is performance than 4 or 64. Except global thresholding, for 16
done in MATLAB 7.0 using a computer with Intel Core 2 Duo patterns Haar transform based 16-patterns have given better
Processor T8100 (2.1GHz) and 2 GB RAM. The CBIR performance. Except local thresholding, for 64 pattern Kekre
techniques are tested on the Wang image database [18] of transform has shown better performance.
1000 variable size images spread across 11 categories of
human being, animals, natural scenery and manmade things,
etc. The categories and distribution of the images is shown in
table 2.
Table 2. Image Database: Category-wise Distribution

Category Tribes Buses Beaches
No. of
85 99 99
Images
Category Horses Mountains Airplanes
No. of
99 61 100
Images
Category Dinosaurs Elephants Roses
No. of Figure 5. Crossover points of precision-recall for proposed texture pattern
99 99 99
Images based CBIR methods with respect to the considered image transforms
Category Monuments Sunrise
No. of
99 61
Images

To assess the retrieval effectiveness, we have used the
precision and recall as statistical comparison parameters [4,5]
for the proposed CBIR techniques. The standard definitions
for these two measures are given by the equations 17 and 18.
Number _ of _ relevant _ images _ retrieved
Pr ecision  (17)
Total _ number _ of _ images _ retrieved
Number _ of _ relevant _ images _ retrieved
Re call  (18)
Total _ number _ of _ relevent _ images _ in _ database

Figure 6. Crossover points of precision-recall for proposed texture pattern
based CBIR methods with respective image map thresholding techniques
VI. RESULTS OF CBIR USING TEXTURE PATTERNS
Figure 7 gives crossover points of precision-recall for
The crossover point of precision-recall plays very proposed texture pattern based CBIR methods with
important role in performance comparison of image retrieval corresponding number of patterns considered. Here 16 texture
methods, higher crossover point value indicates better image patterns have shown better performance than 4 or 64 texture
retrieval. The crossover points of average precision–recall patterns. In case of 4-texture patterns all transforms have
values of firing 55 queries on image database for proposed shown same performance (because of the 2x2 transform
texture pattern based image retrieval methods are computed matrices for all transforms are alike). In 16 patterns except
and plotted in figures 5, 6 and 7. Figure 5 gives crossover global thresholding Haar transform performs better. In case of
points of precision-recall for proposed texture pattern based 64 patterns better performance is given by Kekre transform
CBIR methods with respect to the considered image except local thresholding. The Haar 16 pattern based CBIR
transforms alias Walsh, Haar and Kekre. In all transform with intermediate 4 thresholding has shown best performance
texture pattern based CBIR methods except Walsh local among all CBIR variations considered. In texture pattern
thresholding, 16 patterns have consistently performed well. based CBIR, image retrieval using the Haar 16 patterns with
Also for all transforms intermediate-4 thresholding has given intermediate 4 thresholding has given best performance with
better performance in all 4, 16 and 64 texture patterns. The precision-recall crossover point value 0.461524. The second
crossover points of precision-recall for proposed texture best performance with precision-recall crossover point value

80 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
0.45 is given by CBIR with Kekre 16 patterns with Table 4. Performance Comparison of Number of Texture Patterns considered
for image retrieval using Texture Patterns
intermediate 4 thresholding, the next in the performance is
image retrieval based on Kekre 16 patterns with global Number of Average
Comparative
thresholding with crossover point value 0.4489 followed by Texture Crossover
Performance
Haar 16 patterns with global thresholding and crossover point Patterns Point Value
value 0.44834. Best 16 Pattern 0.43466
Second Best 64 Pattern 0.415209
Poorest 4 Pattern 0.412495

Table. 5 Performance Comparison of Thresholding method used to prepare
image maps for image retrieval using Texture Patterns

Average
Comparative Thresholding
Crossover
Performance Method
Point Value
Best Intermediate 4 0.43597
Second Best Global 0.42180
Third Best Intermediate 9 0.41363
Figure 7. Crossover points of precision-recall for proposed texture pattern Poorest Local 0.41171
based CBIR methods with corresponding number of patterns considered

VII. PERFORMANCE COMPARISION OF VARIANTS IN TEXTURE Total four varied thresholding methods are considered for
PATTERN BASED CBIR METHODS image maps preparation for image retrieval using texture
patterns whose performance comparison is given in table 5 by
The novel image retrieval methods using texture patterns
means of average precision-recall crossover point values of
are presented in this section. Here in all 36 variations of
texture based CBIR variants using respective thresholding
proposed image retrieval methods with texture patterns are
method. Intermediate 4 thresholding has been proven better.
proposed using three image transforms (Haar, kekre & Walsh),
The performance ranking for thresholding methods used in
three types of texture patterns (16, 64 & 4) and four ways of
proposed CBIR with texture patterns, starting with the best can
thresholding used to prepare image maps (intermediate-4,
be given as intermediate 4, global, intermediate 9 and local.
global, intermediate-9 & local). The average of precision-
recall crossover point values for respective variation is VIII. REFERENCES
considered for the performance ranking of these variations. [1] Dr. H.B.Kekre, Sudeep D. Thepade, Varun K. Banura, “Augmentation
Three image transforms namely Walsh, Haar and Kekre are of Colour Averaging Based Image Retrieval Techniques using Even
considered to generate texture patterns. From the results after part of Images and Amalgamation of feature vectors”, International
experimentation it is found that the Haar transform is showing Journal of Engineering Science and Technology (IJEST), Volume 2,
Issue 10, (ISSN: 0975-5462) Available online at http://www.ijest.info
best performance followed by Kekre transform and then
[2] Dr. H.B.Kekre, Sudeep D. Thepade, Varun K. Banura, “Amelioration
Walsh transform in proposed CBIR methods as indicated by of Colour Averaging Based Image Retrieval Techniques using Even
average precision-recall crossover point values of texture and Odd parts of Images”, International Journal of Engineering Science
based CBIR variants using respective image transform given and Technology (IJEST), Volume 2, Issue 9, (ISSN: 0975-5462)
in table 3. The number of texture patterns considered here are Available online at http://www.ijest.info.
4, 16 and 64. The 16 texture patterns have shown better [3] Dr. H.B.Kekre, Sudeep D. Thepade, Akshay Maloo, “Query by Image
Content Using Colour Averaging Techniques”, International Journal of
performance in CBIR using texture patterns. The 64 texture Engineering Science and Technology (IJEST), Volume 2, Issue 6,
patterns have given second best performance followed by 4 2010.pp.1612-1622 (ISSN: 0975-5462) Available online at
texture patterns with poorest performance as per the average http://www.ijest.info.
precision-recall crossover point values of texture based CBIR [4] Dr. H.B.Kekre, Sudeep D. Thepade, “Boosting Block Truncation
variants using respective number of texture patterns given in Coding using Kekre‟s LUV Color Space for Image Retrieval”, WASET
International Journal of Electrical, Computer and System Engineering
table 4. (IJECSE), Volume 2, Number 3, pp. 172-180, Summer 2008. Available
Table 3. Performance Comparison of Image Transforms used in image online at http://www.waset.org/ijecse/v2/v2-3-23.pdf
retrieval using Texture Patterns [5] Dr. H.B.Kekre, Sudeep D. Thepade, “Image Retrieval using
Augmented Block Truncation Coding Techniques”, ACM International
Average Conference on Advances in Computing, Communication and Control
Comparative Image
Crossover (ICAC3-2009), pp. 384-390, 23-24 Jan 2009, Fr. Conceicao Rodrigous
Performance Transform
Point Value College of Engg., Mumbai. Is uploaded on online ACM portal.
Best Haar 0.42320 [6] Dr. H.B.Kekre, Sudeep D. Thepade, “Scaling Invariant Fusion
Second Best Kekre 0.41998 of Image Pieces in Panorama Making and Novel Image
Blending Technique”, International Journal on Imaging (IJI),
Poorest Walsh 0.41916 www.ceser.res.in/iji.html, Volume 1, No. A08, pp. 31-46, Autumn
2008.

81 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
[7] Hirata K. and Kato T. “Query by visual example – content-based image [23] Dr. H.B.Kekre, Sudeep D. Thepade, “Color Based Image Retrieval
retrieval”, In Proc. of Third International Conference on Extending using Amendment Block Truncation Coding with YCbCr Color Space”,
Database Technology, EDBT‟92, 1992, pp 56-71 International Journal on Imaging (IJI), Volume 2, Number A09,
[8] Dr. H.B.Kekre, Sudeep D. Thepade, “Rendering Futuristic Image Autumn 2009, pp. 2-14. Available online at www.ceser.res.in/iji.html.
Retrieval System”, National Conference on Enhancements in Computer, [24] Dr. H.B.Kekre, Tanuja Sarode, Sudeep D. Thepade, “Color-Texture
Communication and Information Technology, EC2IT-2009, 20-21 Mar Feature based Image Retrieval using DCT applied on Kekre‟s Median
2009, K.J.Somaiya College of Engineering, Vidyavihar, Mumbai-77. Codebook”, International Journal on Imaging (IJI), Volume 2, Number
[9] Minh N. Do, Martin Vetterli, “Wavelet-Based Texture Retrieval Using A09, Autumn 2009,pp. 55-65. Available online at
Generalized Gaussian Density and Kullback-Leibler Distance”, IEEE www.ceser.res.in/iji.html (ISSN: 0974-0627).
Transactions On Image Processing, Volume 11, Number 2, pp.146-158, [25] Dr. H.B.Kekre, Sudeep D. Thepade, Akshay Maloo “Performance
February 2002. Comparison for Face Recognition using PCA, DCT &WalshTransform
[10] B.G.Prasad, K.K. Biswas, and S. K. Gupta, “Region –based image of Row Mean and Column Mean”, ICGST International Journal on
retrieval using integrated color, shape, and location index”, Graphics, Vision and Image Processing (GVIP), Volume 10, Issue II,
International Journal on Computer Vision and Image Understanding Jun.2010, pp.9-18, Available online
Special Issue: Colour for Image Indexing and Retrieval, Volume 94, http://209.61.248.177/gvip/Volume10/Issue2/P1181012028.pdf..
Issues 1-3, April-June 2004, pp.193-233. [26] Dr. H.B.Kekre, Sudeep D. Thepade, “Improving the Performance of
[11] Dr. H.B.Kekre, Sudeep D. Thepade, “Creating the Color Panoramic Image Retrieval using Partial Coefficients of Transformed Image”,
View using Medley of Grayscale and Color Partial Images ”, WASET International Journal of Information Retrieval, Serials Publications,
International Journal of Electrical, Computer and System Engineering Volume 2, Issue 1, 2009, pp. 72-79 (ISSN: 0974-6285)
(IJECSE), Volume 2, No. 3, Summer 2008. Available online at [27] Dr. H.B.Kekre, Sudeep D. Thepade, Archana Athawale, Anant Shah,
www.waset.org/ijecse/v2/v2-3-26.pdf. Prathmesh Verlekar, Suraj Shirke, “Performance Evaluation of Image
[12] Stian Edvardsen, “Classification of Images using color, CBIR Distance Retrieval using Energy Compaction and Image Tiling over DCT Row
Measures and Genetic Programming”, Ph.D. Thesis, Master of science Mean and DCT Column Mean”, Springer-International Conference on
in Informatics, Norwegian university of science and Technology, Contours of Computing Technology (Thinkquest-2010), Babasaheb
Department of computer and Information science, June 2006. Gawde Institute of Technology, Mumbai, 13-14 March 2010, The paper
will be uploaded on online Springerlink.
[13] Dr. H.B.Kekre, Tanuja Sarode, Sudeep D. Thepade, “DCT Applied to
Row Mean and Column Vectors in Fingerprint Identification”, In [28] Dr. H.B.Kekre, Tanuja K. Sarode, Sudeep D. Thepade, Vaishali
Proceedings of International Conference on Computer Networks and Suryavanshi,“Improved Texture Feature Based Image Retrieval using
Security (ICCNS), 27-28 Sept. 2008, VIT, Pune. Kekre‟s Fast Codebook Generation Algorithm”, Springer-International
Conference on Contours of Computing Technology (Thinkquest-2010),
[14] Zhibin Pan, Kotani K., Ohmi T., “Enhanced fast encoding method for
Babasaheb Gawde Institute of Technology, Mumbai, 13-14 March
vector quantization by finding an optimally-ordered Walsh transform
2010, The paper will be uploaded on online Springerlink.
kernel”, ICIP 2005, IEEE International Conference, Volume 1, pp I -
573-6, Sept. 2005. [29] Dr. H.B.Kekre, Tanuja K. Sarode, Sudeep D. Thepade, “Image
Retrieval by Kekre‟s Transform Applied on Each Row of Walsh
[15] Dr. H.B.Kekre, Sudeep D. Thepade, “Improving „Color to Gray and
Transformed VQ Codebook”, (Invited), ACM-International Conference
Back‟ using Kekre‟s LUV Color Space”, IEEE International Advanced
and Workshop on Emerging Trends in Technology (ICWET
Computing Conference 2009 (IACC‟09), Thapar University, Patiala,
2010),Thakur College of Engg. And Tech., Mumbai, 26-27 Feb 2010,
INDIA, 6-7 March 2009. Is uploaded at online at IEEE Xplore.
The paper is invited at ICWET 2010. Also will be uploaded on online
[16] Dr. H.B.Kekre, Sudeep D. Thepade, “Image Blending in Vista Creation ACM Portal.
using Kekre's LUV Color Space”, SPIT-IEEE Colloquium and
[30] Haar, Alfred, “ZurTheorie der orthogonalenFunktionensysteme”.
International Conference, Sardar Patel Institute of Technology, Andheri,
(German), MathematischeAnnalen, volume 69, No. 3, 1910, pp. 331–
Mumbai, 04-05 Feb 2008.
371.
[17] Dr. H.B.Kekre, Sudeep D. Thepade, “Color Traits Transfer to
[31] Charles K. Chui, “An Introduction to Wavelets”, Academic Press, 1992,
Grayscale Images”, In Proc.of IEEE First International Conference on
San Diego, ISBN 0585470901.
Emerging Trends in Engg. & Technology, (ICETET-08), G.H.Raisoni
COE, Nagpur, INDIA. Uploaded on online IEEE Xplore. [32] Dr. H. B. Kekre, Sudeep D. Thepade, Varun K. Banura, “Query by
Image Texture Pattern content using Haar Transform Matrix and Image
[18] http://wang.ist.psu.edu/docs/related/Image.orig (Last referred on 23
Bitmaps”, Invited at ACM International Conference and Workshop on
Sept 2008)
Emerging Trends in Technology (ICWET 2011), TCET, Mumbai, 25-
[19] Dr. H.B.Kekre, Sudeep D. Thepade, “Using YUV Color Space to Hoist 26 Feb 2011.
the Performance of Block Truncation Coding for Image Retrieval”,
[33] Dr. H. B. Kekre, Sudeep D. Thepade, “Image Retrieval using Non-
IEEE International Advanced Computing Conference 2009 (IACC‟09),
Involutional Orthogonal Kekre‟s Transform”, International Journal of
Thapar University, Patiala, INDIA, 6-7 March 2009.
Multidisciplinary Research and Advances in Engineering (IJMRAE),
[20] Dr. H.B.Kekre, Sudeep D. Thepade, Archana Athawale, Anant Shah, Ascent Publication House, 2009, Volume 1, No.I, pp 189-203, 2009.
Prathmesh Verlekar, Suraj Shirke,“Energy Compaction and Image Abstract available online at www.ascent-journals.com
Splitting for Image Retrieval using Kekre Transform over Row and
[34] Dr. H. B. Kekre, Sudeep D. Thepade, Varun K. Banura, “Image
Column Feature Vectors”, International Journal of Computer Science
Retrieval using Texture Patterns generated from Walsh-Hadamard
and Network Security (IJCSNS),Volume:10, Number 1, January 2010,
Transform Matrix and Image Bitmaps”, (Selected) Springer
(ISSN: 1738-7906) Available at www.IJCSNS.org.
International Conference on Technology Systems and Management
[21] Dr. H.B.Kekre, Sudeep D. Thepade, Archana Athawale, Anant Shah, (ICTSM 2011), MPSTME and DJSCOE, Mumbai, 25-27 Feb 2011.
Prathmesh Verlekar, Suraj Shirke, “Walsh Transform over Row Mean
[35] Dr. H.B.Kekre, Sudeep D. Thepade, Shrikant Sanas, “Improving
and Column Mean using Image Fragmentation and Energy Compaction
Performance of multileveled BTC based CBIR using Sundry Color
for Image Retrieval”, International Journal on Computer Science and
Spaces”, CSC International Journal of Image Processing (IJIP), Volume
Engineering (IJCSE),Volume 2S, Issue1, January 2010, (ISSN: 0975–
4, Issue 6, Computer Science Journals, CSC Press,
3397). Available online at www.enggjournals.com/ijcse.
www.cscjournals.org
[22] Dr. H.B.Kekre, Sudeep D. Thepade, “Image Retrieval using Color-
[36] Dr. H.B.Kekre, Sudeep D. Thepade, Akshay Maloo “Performance
Texture Features Extracted from Walshlet Pyramid”, ICGST
Comparison of Image Retrieval Using Fractional Coefficients of
International Journal on Graphics, Vision and Image Processing
Transformed Image Using DCT, Walsh, Haar and Kekre‟s Transform”,
(GVIP), Volume 10, Issue I, Feb.2010, pp.9-18, Available online
CSC International Journal of Image Processing (IJIP), Volume 4, Issue
www.icgst.com/gvip/Volume10/Issue1/P1150938876.html

82 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 09, No.03, 2011
2, pp 142-157, Computer Science Journals, CSC Press, Management and Engineering, SVKM‟s NMIMS University,
www.cscjournals.org Vile Parle(w), Mumbai, INDIA. He is member of
International Association of Engineers (IAENG) and
International Association of Computer Science and
AUTHORS PROFILE
Information Technology (IACSIT), Singapore. He has been
on International Advisory Board of many International
Dr. H. B. Kekre has received B.E. (Hons.) in Telecomm. Conferences. He is Reviewer for many reputed International
Engineering. from Jabalpur University in 1958, M.Tech Journals. His areas of interest are Image Processing and
(Industrial Electronics) from IIT Bombay in 1960, Computer Networks. He has more than 100 research papers
M.S.Engg. (Electrical Engg.) from University of Ottawa in in National/International Conferences/Journals to his credit
1965 and Ph.D. (System Identification) from IIT Bombay with a Best Paper Award at International Conference
in 1970 He has worked as Faculty of Electrical Engg. and SSPCCIN-2008, Second Best Paper Award at ThinkQuest-
then HOD Computer Science and Engg. at IIT Bombay. For 2009 National Level paper presentation competition for
13 years he was working as a professor and head in the faculty, Best Paper Award at Springer International
Department of Computer Engg. at Thadomal Shahani Conference ICCCT-2010 and second best project award at
Engineering. College, Mumbai. Now he is Senior Professor Manshodhan 2010.
at MPSTME, SVKM‟s NMIMS University. He has guided
17 Ph.Ds, more than 100 M.E./M.Tech and several
B.E./B.Tech projects. His areas of interest are Digital Signal Varun K. Banura is currently pursuing B.Tech. (CE) from
processing, Image Processing and Computer Networking. He MPSTME, NMIMS University, Mumbai. His areas of
has more than 320 papers in National / International interest are Image Processing and Computer Networks. He
Conferences and Journals to his credit. He was Senior has 07 research papers in International Conferences/Journals
Member of IEEE. Presently He is Fellow of IETE and Life to his credit.
Member of ISTE Recently ten students working under his
guidance have received best paper awards and two have been
conferred Ph.D. degree of SVKM‟s NMIMS University.
Currently 10 research scholars are pursuing Ph.D. program
under his guidance.

Sudeep D. Thepade has Received B.E.(Computer) degree
from North Maharashtra University with Distinction in 2003.
M.E. in Computer Engineering from University of Mumbai
in 2008 with Distinction, currently pursuing Ph.D. from
SVKM‟s NMIMS, Mumbai. He has about than 07 years of
experience in teaching and industry. He was Lecturer in
Dept. of Information Technology at Thadomal Shahani
Engineering College, Bandra(w), Mumbai for nearly 04
years. Currently working as Associate Professor in Computer
Engineering at Mukesh Patel School of Technology

83 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011

A Generic Rule-Based Agent for Monitoring Temporal Data Processing

S. Laban A.I. El-Desouky A. S. ElHefnawy
International Data Centre (IDC) Computer and Systems Department, Information Technology
Comprehensive Nuclear Test-Ban Faculty of Engineering, Mansoura Department, Faculty of Computer
Treaty Organization (CTBT), University, Mansoura, Egypt & Information, Mansoura
Vienna, Austria* University, Mansoura, Egypt
shaban.laban@ctbto.org

Abstract—Most of the current real-time monitoring tools are task restrictions. The agent proposes a unified and flexible format
specific, lacking alerts capabilities, inflexible, and consuming for representing the temporal data and objects (intervals) as
many of the organization resources in maintaining and adding well as their different attributes. Also, the agent is using
newer monitored objects. This paper introduces the design and customized rules for workflow monitoring and generating
implementation of a generic rule-based agent model that different exception reports and alerts when necessary.
minimizes the previous limitations and restrictions. The proposed
intelligent agent is using dedicated rules for workflow monitoring
and generating alerts as well as exception reports to the
The overall structure of the system that the proposed agent
operators. A unified data model is proposed to reduce the
irregularity and complexity of the monitored data and objects. is modeled is explained in Section 2. In Section 3, detailed
The suggested rule-based monitoring agent is generic, architecture and design of the rule-based agent are presented.
autonomous, configurable, and platform-independent. Real-time practical implementation of the proposed system is
illustrated in Section 4. Section 5 evaluates the performance of
Keywords- rule-based; monitoring; workflow; agent. the suggested agent and presents the results of the comparison
with the available tools. Section 6 concludes this paper by
I. INTRODUCTION summarizing the contributions of the proposed approach and
The workflow of real-time data processing systems consist discussing future directions.
generally of series or sequence of complex stages and
processes [1]. In every stage, the states of the different objects
or elements, as well as their internal attributes, are dynamically II. THE OVERALL FRAMEWORK
changed during its different steps. Generally, real-time systems In order to achieve the maximum scalability, efficiency,
use monitoring tools to increase their productivity and and robustness of the overall system, the web-enabled
efficiency by detecting anomalies, potential workflow failures monitoring approach is structured into three main different
and tracing workflow progress of the different processes [2-4]. modules as shown in Fig. 1. The arrows in the graph indicate
The operators of the real-time systems use several instances of the direction of data flow between the different modules. The
the monitoring tools to supervise, control, monitor workflow dashed line indicates the direct data flow for intranet
progress and trace the states of the different objects and monitoring agents.
resources during the life cycle of monitored systems.

Most of the traditional monitoring tools have many Real-time Data Rule-Based
limitations. Such monitoring tools are usually platform Processing & Web server/ Monitoring
Services Agent
dependent, task specific, inflexible, and having limited Repositories
resources management. Also, the monitoring tools consume
most of the organization resources, need long time, more Figure 1. Overall framework
human resources to maintain or configure newer monitored The first module comprises a set of knowledge-based
objects. Moreover, usually they lack intelligence decision repositories as well as intelligent scheduled or continuous
support. In addition, the monitoring tools are not available to running programs. These programs act as data collectors,
remote users and are not portable. Furthermore, the monitored processors, and producers. They regularly collect states of
data have no standard format or unified structures which objects from the different data sources of the monitored system.
complicate building those monitoring tools. Then, they infer the monitored data and classify the status of
The aim of this paper is to present an approach for the different objects as well as their necessary
implementing a platform-independent and task-unspecific rule- attributes/properties from the different stages of the monitored
based agent model that minimizes the previous limitations and system. Finally, these programs store the data into dedicated
*
Disclaimer: The views expressed on this paper are those of the authors and
do not necessarily reflect the view of the CTBTO.

84 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
data repositories either in an Extensible Markup Language A. Unified Data Representation Model
(XML) format or in a unified standard data format [8,9].
In the monitoring agent, and in order to reduce data
irregularity and complexity, the monitored objects are
The second module consists of any ordinary web server and represented and modeled as temporal (interval-based) elements
a set of CGI programs or services (act as intermediate data [15]. Each interval-based element, in our case
producers) in charge to read the data generated from the first IntervalElementInfo class, is composed of a set of well defined
module, prepare the browser configuration data and a user fields or slots and array of optional attributes (AttributeInfo
specific views and to send all the information to the monitoring class). The simplified UML class diagram and composition
agents in Hypertext Markup Language (HTML) using relationship are shown in Fig. 3.
Hypertext Transfer Protocol (HTTP) [10,11].
IntervalElementInfo AttributeInfo
ElementId:int Name:String
ElementClass:String Value:String
The third module is our proposed monitoring agent (data ElementName:String Alias:String
consumers). This agent consists of a set of Java classes or ElementState:String color:Color
programs that can be run either from the web browsers for all ElementEpocStart:int ….
known platforms as a Java applet or in standalone mode [12- ElementEpocEnd:int getName():String
14]. ElementColor:Color getValue():String
Attributes[]:AttributeInfo getAlias():String
….. setColor(Color)
getColor():Color
III. THE PROPOSED MONITORING AGENT MODEL addAttribute(AttributeInfo)
setElementClass(String)
In this paper, the rule-based techniques and agent …
programming concepts are applied to workflow monitoring in getAttribute():AttributeInfo
order to provide more flexibility and intelligence to the getElementClass():String
monitoring process [18]. The agent reads regularly the ….
getState():String
necessary information from the web server using the HTTP setState(String)
protocol. Then, the agent aggregate, sort, infer, and finally Figure 3. Class diagram and composition relationship
display the monitored data accordingly in the web browser.
The agent is using a dedicated customized rule engine for The implementation of such unified data format is
monitoring purposes as well as generating exceptional reports necessary for the monitoring agent model in order to support
and early warning alerts to the operators. The agent is able to new generations of portable digital assistance devices such as
communicate with its user/operator. It reads the user updated iPhone, embedded systems, as well as other similar devices.
configuration parameters and reacts with user requests and
needs. The internal structure including the main functions of
the monitoring agent is shown in Fig. 2. B. Configuration and Control
The agent configuration is carried out either during
program loading (default configuration) or while the agent is
running (user preferences). During program initialization, the
Requests Configuration/ parameters are fed to the agent using the HTML tag
Monitored <APPLET> or fed from a configuration file if the agent is
Data
running in a standalone mode.
External interface During run time, the user can change most of loaded
Configuration parameters. This is done by selecting either “Sort/Filter
& Control Update Update Viewed Data” tab, “View/Update Rules” tab or
Agent knowledge- “Configuration” tab. The configuration and control parameters
base are classified into four main categories as follows:
Working Memory
(facts, reports,
alerts) Monitoring rules The first category contains the connection parameters that
are necessary for identifying and connecting to the specified
HTTP server.
GUI components Inference engine
The second category contains the parameters that are
needed to read, parse, and display the received data from the
Figure 2. The internal structure of the monitoring agent model server.

85 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
While, the third category contains the necessary parameters
that are needed to set up the graphical display of the applet
itself.

Finally, the last category contains the parameters needed to
read, parse, map the various rules definitions, generate alerts
and exception reports and display the various states of the
interval.

C. Rule Engine
The agent uses a time-dependent rule engine for states
identification and classification. The engine is a forward-
chaining rule engine based on JRuleEngine [16]. Default rules
definition is loaded at start up time from the HTTP server.
Simplified rule syntax can be described as follows:

RULE-n [description]
[SALIENCE salience]
IF Figure 4. Agent rules window

IntervalElement.Property Operator State1 [State2]
[IntervalElement.Property Operator state1 [State2] ]
THEN
set State Identifier <color code>
[THEN generate alert for Operator]

Due to the nature of the monitored temporal data and the
agent, currently one rule will be fired per matched interval for
state setting. The rule with higher salience will have high
priority over the rule with lower salience. The agent is using by
default depth strategy for conflict resolution. This means that
newly activated rules are placed above all rules of the same
salience. Additionally, an exceptional conflict report entry will
be available to the operator for conflicting rules.
The agent graphical rule implementation of the proposed
rule syntax is illustrated in Fig.4. The rules can be dynamically
inserted during run-time as well as adding new conditions for
any existing rule. The rules set are displayed in an editable
table that could be altered by the user. In addition, the user can
change directly the state-color relationship, by pressing the
color itself. A pop up dialog will appear and the user can select
his new color and hit the button “OK”. Also, check box is
available for every rule to allow the agent to generate alerts
automatically when rule matched. To activate the overall
changes, the user needs to press the button “Update Workflow”
in order to force the agent to immediately reprocess the new
rules definition. Fig. 5 shows a snapshot of an alert window. Figure 5. Example of an alerts window

86 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
D. Main Graphical User Interface Components to view older intervals. Another vertical scrollbar is linked to
Fig. 6 shows the different components of the monitoring the two previous panels. It enables the user to view the hidden
agent. The main component of the agent is the tabbed pane. data if the number of pairs is greater than the number of objects
Currently, it contains six child panels. The child panels include of the “Object Property” panel. A right mouse click on any
the graphical panel, the sort panel, the rules panel, the brick will pop up a window that displays all available
exceptional reports panel, the alerts panel and the configuration attributes/information for this particular interval. Also, a
panel. middle mouse click outside the bricks will pop up a menu that
allows the user to change the current view. Finally, a zoom or
The graphical panel consists of many different widgets or time expander scrollbar is created to allow the user to expand
components to support monitoring the data over time. The first or zoom into the viewed area to display more detailed
panel is the “Object Property” panel. Each “Object Property” information about those intervals.
pair is displayed in a separate row. The main component is the
data display panel. This panel displays the status of the
interval data in colored horizontal rows. The rows are
displayed adjacent to their corresponding “class name” objects.
Each horizontal row is composed of time interval columns or
bricks. The bricks are colored according to the status of that
interval where the mapping between states and colors are
defined and configured by specified rules. A dedicated
horizontal scrollbar is linked to that panel that allows the user

Tabbed pane used for switching between
Current time
different agent panels.

“Object Property” Panel
Object &
attributes
pop up
Window,
alerts are in
“Object Property” scroll

red

Zoom scroll bar Interval time scroll bar

Figure 6. Agent main graphical components

87 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
IV. IMPLEMENTATION The results of the comparison between the proposed agent
The International Data Centre (IDC) of the Preparatory and the other available monitoring tools are summarized in
Commission for the Comprehensive Nuclear-Test-Ban Treaty Table I. If we compare the proposed system with the other
Organization (CTBTO) is receiving and processing, in near real tools on the basis of N different instances, it is clear that the
time, data from all the International Monitoring System (IMS) database and/or resources utilization have been decreased from
stations or facilities via a dedicated Global Communications N to 1 per monitored system. Also, the response time of
Infrastructure (GCI) [17]. Monitoring IMS data includes proposed rule-based monitoring agent is faster time than the
receiving, forwarding, automatic processing, interactive current tools. Moreover, it is dynamic and simpler in
review, generation and distribution of IDC scientific bulletins configuration, integration and implementation. Furthermore,
and reports, and archiving of the data. the proposed system is portable, dynamically configurable, and
generic. Finally, the proposed system is more advantageous
than the other tools because of its web-oriented characteristics
as well as alerts and reports capabilities.
The proposed prototype agent has been implemented in
monitoring the real-time processing of the data received TABLE I. COMPARISON RESULTS
continuously from the IMS stations, including radionuclide and
meteorological data, or similar temporal data. The monitored No. Attribute Other Proposed
data are from different technologies and different Available Agent
characteristics IDC seismic data processing and pipeline, Tools
radionuclide data processing and atmospherics data processing.
A recent real-time snapshot of the proposed agent is shown in 1 Response Time Fast Faster
Fig. 7. 2 Flexibility Hard Easy
3 Configurability Simple Simple
4 Portability No Yes
5 Dynamic Not Easy Yes
Configuration
6 Web-Enabled No Yes
7 Task Specific Yes Generic
8 Scalability Not Easy Easy
9 Resources N 1
Utilization
(per monitored
object)
10 Reports Capability N/A Yes
11 Alerts Capability N/A Yes

VI. CONCLUSION AND FUTURE WORK
The proposed lightweight rule-based agent is generic,
portable, and flexible to handle any time interval, and platform
independent. By implementing the proposed agent, status of
the different objects, as well as their attributes or properties,
can be easily monitored either via the organization Intranet, or
remotely using Virtual Private Network (VPN) connections
through the Internet. The agent allows timely generating
exceptional reports and alerts. This will help the operators and
decision making users to act faster on the anomalies that could
Figure 7. Workflow of IDC radionuclide data processing be occurred during the workflow monitoring process.

V. COMPARISON AND RESULTS Future work will be dedicated to create special monitoring
agent, for simulation and analysis purposes. Also, it is planned
More than ten performance attributes have been identified to integrate or enable the use of the proposed monitoring agent
to evaluate the proposed agent. Also, a comparison between the in grid monitoring systems in order to share and distribute
proposed agent and the current available monitoring tools has generated alerts and exceptional reports with the other intranet
been made. Selected datasets from different records, extracted agents
from IDC databases and different critical processes, have been
used to evaluate the performance of the proposed system with ACKNOWLEDGMENT
the existing similar monitoring tools. The authors express their gratitude to thank CTBTO
management for encouraging the research activities. Special

88 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
thanks to the IDC staff and the radionuclide analysts for the [7] M. Klusch, “Information agent technology for the internet: a Survey”,
constructive comments during testing and implementing the Data & Knowledge Engineering, vol. 36, 2001, 2001.
proposed agent. [8] Extensible Markup Language, http://www.xml.org/
[9] T. Bray et al., Extensible Markup Language (XML): W3C (World-Wide
REFERENCES Web Consortium), http://www.w3c.org/.
[10] The Internet Society (ISOC), http://www.isoc.org/.
[1] F. J. Kurfess and D. P. Shah, "Monitoring distributed process with [11] Request For Comments (RFC) Editor, http://www.rfc-editor.org/.
intelligent agents," In: Proceedings of IEEE Conference and Workshop [12] Java virtual machine, http://java.sun.com/.
on Engineering of Computer-Based Systems, 1998. [13] J.E. Smith and R.Nair., “The architecture of virtual machines”, IEEE
[2] J.-J. Jeng, J. Schiefer, and H. Chang, "An agent-based architecture for Computer, Volume 38(Issue 5):Pages 32–38, May 2005.
analyzing business processes of real-time enterprises," Proceedings of [14] S. Orlando, and S. Russo, “Java virtual machine for dependability
the IEEE 7th International Conference on Enterprise Distributed Object benchmarking”, Proceedings of the Ninth IEEE International
Computing, pp. 86, 2003. Symposium on Object and Component-Oriented Real-Time Distributed
[3] I. Seilonen, T. Pirttioja, A. Halme, K. Koskinen, and A. Pakonen, Computing (ISORC'06),2006,pp. 433-440.
,”Indirect process monitoring with constraint handling agents,” 4th [15] Bradshaw, J.M., Software agents. San Francisco, CA, USA: AAAI
International IEEE Conference on Industrial Informatics, Aug. 2006 , Press/MIT Press, 1997.
pp.1323 – 1328. [16] The Java Rule Engine, http://jruleengine.sourceforge.net/.
[4] L. Bunch, M. Breedy, J. M. Bradshaw, M. Carvalho, N. Suri, A. Uszok, J. [17] Preparatory Commission for the Comprehensive Nuclear-Test-Ban Treaty
Hansen, M. Pechoucek, V.Marik, “Software agents for process (CTBT) Organization, http://www.ctbto.org/.
monitoring and notification,” ACM Symposium.
[18] S. Russell and P. Norvig, Artificial Intelligence, A Modern Approach,
[5] A. M. Shaw, S. Yada, “A comprehensive agent-based architecture for Prentice Hall, 1995.
intelligent information retrieval in a distributed heterogeneous
environment”, Decision Support Systems, vol. 32, 2002, pp. 401-415.
[6] R. Jennings, “An agent-based approach for building complex software
systems”, Communications of the ACM, Vol. 44, No. 4, 2001,pp. 35-41.

89 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

A New Approach for Model based Gait Signature
Extraction
Mohamed Rafi Shanawaz Ahmed J Md. Ekramul Hamid R.S.D Wahidabanu
Dept. of CS&E Dept. of Computer Science Dept. of Computer Network Engg. Dept. of E&C
HMS Institute of Tech. College of Computer Science College of Computer Science Government college of Engg
Tumkur, Karnataka, India King Khalid University, KSA King Khalid University, Abha, KSA Salem, Tamil Nadu, India.
mdrafi2km@yahoo.com jshanawaz@gmail.com ekram_hamid@yahoo.com drwahidabanu@gmail.com

Abstract— Identifying individuals for security purposes are Gait is defined as, “A particular way or manner of moving
becoming essential now-a-day. Gait recognition aims essentially on foot”. Early psychological studies into gait by Murray [2],
to address this problem by identifying people at a distance based suggested that gait is a unique personal characteristic, with
on the way they walk. In this paper, a model is proposed for gait cadence and is cyclic in nature. Gait recognition as a
signature extraction consists of gait capture, segmentation and physiological biometric technique has become popular in
feature extraction steps. We use Hough transform technique that recent times. Gait as a biometric can be seen as advantageous
helps to read all parameters which are used to generate gait [3] over other forms of biometric identification techniques for
signatures that may result a better gait recognition rate. In the it is unobtrusive, can be captured at a distance, does not require
preprocessing steps, the picture frames taken from video
high quality images and it is difficult to disguise. The first
sequences are given as input to Canny edge detection algorithm
to detect edges of the image by extracting foreground from
scientific article on animal walking gaits has been written
background also it reduces the noise using Gaussian filter. The 350BC by Aristotle [4]. He observed and described different
output from edge detection is an input to the Hough transform to walking gaits of bipeds and quadrupeds and analyzed why all
extract gait parameters. We have used five parameters to animals have an even number of legs. Recognition approaches
successfully extract gait signatures. It is observed that when the to gait were first developed in the early 1990s and were
camera is placed at 90 and 270 degrees, all the parameters used evaluated on smaller databases and showed promise. DARPA’s
in the proposed work are clearly visible. Then using Hough Human ID at a Distance program [5] then collected a rich
transform, a clear line based model is designed to extract gait variety of data and developed a wide variety of technique and
signatures. The utility of the model is tested on a variety of body showed not only that gait could be extended to large databases
and stride parameters recovered in different viewing conditions and could handle covariate factors. Since the DARPA program,
on a database consisting of 15 to 20 subjects walking at both an research has continued to extend and develop technique, with
angled and frontal-parallel view with respect to the camera, both especial consideration of practical factors such as feature
indoors and outdoors and find the method to be highly successful. potency.
Keywords- Biometric, Gait signature extraction, Hough In Silhouette Analysis-Based Gait Recognition for Human
Transform, Canny Edge detection, Gaussian filter Identification [6] a combination of background subtraction
procedure and a simple correspondence method was used to
segment and track spatial silhouettes of an image, but this
I. INTRODUCTION method generates more noise which leads to poor gait signature
extraction. Therefore the rate of recognition was low. In gait
The demand for automatic human identification system is recognition by symmetry analysis[7], the Generalized
strongly increasing and growing in many important Symmetry Operator was used which locates features according
applications, especially at a distance and it has recently gained to their symmetrical properties rather than relying on the
great interest from the pattern recognition and computer vision borders of a shape or on general appearance and hence does not
researchers for it is widely used in many security-sensitive require the shape of an object to be known in advance. The
environments such as banks, parks and airports. Biometrics is a evaluation was done by masking with a rectangular bar of
new powerful tool for reliable human identification and it different widths: 5, 10 and 15 pixels in each image frame of the
makes use of human physiology or behavioral characteristics test subject and at the same position. The area masked was on
such as face, iris, fingerprints and hand geometry for average 13.2%, 26.3% and 39.5% of the image silhouettes,
identification. However, these biometrics methodologies are respectively. A recognition rate of 100% was obtained for bar
either instructive or restricted to many controlled environments. size of 5 pixels. For a bar width of 10 pixels the test failed as
For example, most face recognition methods are capable of the test subject could not be recognized as subject was
recognizing only frontal or nearly frontal faces, other completely covered in most of the image frames. This suggests
biometrics such as fingerprint and iris are no longer applicable that recognition is likely to be adversely affected when a
when the persons suddenly appears in the surveillance. subject walks behind a vertically placed object. There were also
Therefore, new biometrics recognition methods are strongly other limitations, Mark Ruane Dawson [8], like the legs were
needed in many surveillance applications, especially at a not being tracked to a high enough standard for gait
distance [1]. recognition. The segmentation process leads to a very crude

90 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
model fitting procedure which in turn adversely affects the rate A. Gait Capturing
of recognition. In other method of gait recognition, the subjects At this step the subjects are asked to walk for capturing of
in the video are always walking perpendicular to the camera gait. This is a very important step as the total results depend on
[9]. This would not be the case in real life as people would be the quality of the gait captured. So the care should be taken to
walking at all angles to the video camera. Using of fewer see that the quality of gait capturing is maintained, this step
parameters for gait signatures was another major drawback includes video sequence and XML data store. In our proposed
which has to be addressed. research the following preprocessing steps are carried out
The motivation behind this research is to find whether before segmenting a captured Image
increase in number of gait signature can improve the • [Reading a RGB image]
recognition rate? Improvement over model fitting can give us
better results? What factors affect gait recognition and to what • [Converting an RGB image to Gray Scale]
extent? And what are the critical vision components affecting
gait recognition from video? The objective of this paper is to • [Converting Gray Scale Image to Indexed Image]
explore the possibility of extracting a gait biometric from a The indexed image is the input to the segmentation algorithm
sequence of images of a walking subject without using for further processing. The above preprocesses of an image is
markers. Sophisticated computer vision techniques are shown in figure 2.
developed, aimed to extract a gait signature that can be used for
person recognition. B. Segmentation
In computer vision, segmentation refers to the process of
Using video feeds from conventional cameras and without
partitioning a digital image into multiple segments (sets of
the use of special hardware, implicates the development of a
pixels, also known as super pixels). The goal of segmentation is
marker less body motion capture system. Research in this
to simplify and/or change the representation of an image into
domain is generally based on the articulated-models approach.
something that is more meaningful and easier to analyse. Image
Haritaoglu et al. [10] presented an efficient system capable of
segmentation is typically used to locate objects and boundaries
tracking 2D body motion using a single camera. Amos Y.
(lines, curves, etc.) in images. More precisely, image
Johnson[11] used a single camera with the viewing plane
segmentation is the process of assigning a label to every pixel
perpendicular to the ground plane, 18 subjects walked in an
in an image such that pixels with the same label share certain
open indoor-space at two view angles: a 45◦ path (angle view)
visual characteristics.
toward the camera, and a frontal-parallel path (side-view) in
relation to the viewing plane of the camera. The side-view data The Canny Edge Detection Algorithm:
was captured at two different depths, 3.9 meters and 8.3 meters
from camera. These three viewing conditions are used to The picture frames taken from video sequences are given as
evaluate our multi-view technique. In this research, we use input to Canny edge detection algorithm to detect the edges of
images captured at different views as the image captured from the image frames.
the frontal or perpendicular view does not give required The algorithm consists of 5 steps:
signatures. Segmentation is done on the captured image in
order to extract foreground from back ground using Canny 1. Image Smoothing:
edge detection technique, as the purpose of edge detection in Image smoothing is used to reduce noise within an image.
general is to significantly reduce the amount of data in an The Canny edge detector uses a filter based on the first
image, while preserving the structural properties to be used for derivative of a Gaussian, in the form:
further image processing. In order to obtain the gait model the
output of segmentation is processed using Hough transform,
which is a technique that can be used to isolate features of a
particular shape within an image (1)
2. Finding gradients
II. MODEL FOR GAIT SIGNITURE EXTRACTION
We propose a gait signature extraction model having the The edges of the image is marked where the gradients of
the image has large magnitudes. The Canny algorithm basically
following steps- Picture frame capture, Segmentation, feature
finds edges where the grayscale intensity of the image changes
Extraction which leads to gait signature identification which
the most. These areas are found by determining gradients of the
shown in figure.1. image. First step is to approximate the gradient in the x- and y-
direction respectively by applying the kernels. Then the
gradient magnitudes (also known as the edge strengths) are
determined as an Euclidean distance measure by applying the
law of Pythagoras is given by equation
|G| = SQRT(Gx2 + Gy2) (2)
It is simplified by applying Manhattan distance measure is
given by |G| = |Gx| + |Gy|, where Gx and Gy are the gradients in
Figure1. Components of the proposed model for Gait Signature Extraction
System.
Identify applicable sponsor/s here. (sponsors)

91 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
the x- and y-directions, respectively. The direction of the edges parametric form, the classical Hough transform is most
is determined and stored as given by the equation below commonly used technique for the detection of regular curves
such as lines, circles, ellipses, etc. A convenient equation for
∂ = arctan |Gy|/|Gx| (3) describing a set of lines uses parametric or normal notion:
3. Non-maximum suppression x cosθ + ysin θ = r (4)
In the proposed study, only local maxima are marked as where r is the length of a normal from the origin to this line
edges. The purpose of this step is to convert the “blurred” and θ is the orientation of r with respect to the x-axis. For any
edges in the image of the gradient magnitudes to “sharp” edges. point (x,y) on this line, r and θ are constant.
Basically, this is done by preserving all the local maxima in
the gradient image, and deleting everything else. In an image analysis context, the coordinates of the point(s)
of edge segments (i.e.(xi,yi)) in the image are known and
The algorithm for each pixel in the gradient image: therefore serve as constants in the parametric line equation,
a. Round the gradient direction to nearest 45 degrees, while r and θ are the unknown variables we seek. If we plot the
corresponding to the use of an 8-connected possible (r,θ) values defined by each (xi,yi), points in Cartesian
neighborhood. image space map to curves (i.e. sinusoids) in the polar Hough
parameter space. This point-to-curve transformation is the
b. Compare the edge strength of the current pixel with Hough transformation for straight lines. When viewed in
the edge strength of the pixel in the positive and Hough parameter space, points which are collinear in the
negative gradient direction, i.e., if the gradient Cartesian image space become readily apparent as they yield
direction is north (θ =90 degrees), compare with the curves which intersect at a common (r, θ) point.
pixels to the north and south.
The transform is implemented by quantizing the Hough
c. If the edge strength of the current pixel is largest; parameter space into finite intervals or accumulator cells. As
preserve the value of the edge strength. If not, the algorithm runs, each (xi,yi) is transformed into a discretized
suppress (i.e. remove) the value. (r,θ ) curve and the accumulator cells which lie along this curve
4. Double thresholding are incremented. Resulting peaks in the accumulator array
represent strong evidence that a corresponding straight line
Potential edges are determined by thresholding. exists in the image.
5. Edge tracking by hysteresis The main advantage of the Hough transform technique is
Finally edges are determined by suppressing all edges that that it is tolerant of gaps in feature boundary descriptions and is
are not connected to a very certain (strong) edge as shown in relatively unaffected by image noise. We use this technique to
figure 2 extract lines from the segmented image. The Hough transform
can be used to identify the parameter(s) of a curve which best
fits a set of given edge points. This edge description is obtained
from the Canny edge detector and may be noisy, i.e. it may
contain multiple edge fragments corresponding to a single
whole feature. Furthermore, as the output of an edge detector
defines only where features how many are in an image, the
work of the Hough transform is to determine both what the
features are (i.e. to detect the feature(s) for which it has a
parametric (or other) description) and of them exist in the
image.
Hough Transform Algorithm for Straight Lines:
1. Identify the maximum and minimum values of r and θ
2. Generate an accumulator array A(r, θ) and set all
values to zero
3. For all edge points (xi, yi) in the image
a. Use gradient direction for θ
b. Compute r from the equation
Figure 2: [a] Original Image [b]. RGB to Grayscale [c] Grayscale to Indexed c. Increment A(r, θ) by one
Image [d] Edge Detected Image.
4. For all cells in A(r, θ)
C. Gait Feature Extraction
The Hough transform is a technique which can be used to a. Search for the maximum value of A(r, θ)
isolate features of a particular shape within an image. Because b. Calculate the equation of the line
it requires that the desired features be specified in some

92 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
5. To reduce the effect of noise more than one element
(elements in a neighbourhood) in the accumulator array
are increased.
The edge detected image and the image after Hough
transform are shown in figure 3.
Table 2: Parameters and percentage of clarity when camera placed at 270
degree angle to the subject for frame 1.

Figure 3: Images before and after the Hough Transform
Figure 5. Graphical representation of clarity for frame 1, when camera placed at
III. EXPERIMENTAL RESULTS AND DISCUSSION 270 degree.
One of the most important aspects of this research is to Table 1 and Table 2 show the result. While the camera is
extract the gait signatures for a successful recognition rate. placed at 90 degrees and 270 degrees, it is found that for frame
Below table shows the number of parameters which are used to 1 the clarity for the parameter distance between the legs is
generate a gait signature for different view of a subject(90 higher; the y axis is taken as the reference axis for the subject.
degree and 270 degree).The attempts column shows how many Therefore this can be used to extract gait signatures for better
persons are used to extract the signature. The success column recognition. It is also observed that the parameter right thigh
shows how many of the subjects give successful gait length can also be considered for extraction of gait signature. It
signatures. is also observed that while the camera is placed at 90 degrees
and 270 degrees for frame 2, the clarity for the parameter right
thigh length is higher. Therefore, this can be used to extract
gait signatures for better recognition. While placing camera at
90 degrees and 270 degrees for frame 3, it is found that the
clarity for the parameter left thigh length is higher. Therefore
this can also be used to extract gait signatures for better
recognition.
Table 1: Parameters and percentage of clarity when camera placed at 90 degree CONCLUSIONS
angle to the subject for frame 1.
The presented research has shown that gait signatures can
be extracted in a better way by using Hough transform. When
the camera is placed at 90 and 270 degrees it is found that most
parameters listed in the research are providing us clarity. Since
the lines are clearly visible they can easily be labeled and the
distance and angle between them can be measured accurately.
The proposed research gives best results if the camera is
placed at 90 degrees to the subject and it is recommended that
the subjects must be made to pass through an area which has a
white background because it will help in getting a better gait
signature extraction model. The research achieved 100 percent
clarity if the parameters length of left, right thigh and distance
between the legs are analyzed at 90 degree angle. The
signatures thus extracted can be used to get better gait
recognition rate. In future work it is recommended that the
lines extracted from Hough transform should be labeled by
Figure 4. Graphical representation of clarity for frame 1,When camera placed at using an effective line labeling algorithm to calculate the
90 degree.
angles and distances between the various parameters to get
effective gait recognition.

93 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

REFERENCES
[1] Jiwen Lu A, Erhu Zhang B .”Gait recognition for human identification
based on ICA and fuzzy SVM through multiple views fusion”, School of
Electrical and Electronic Engineering, Nanyang Technological
University, Nanyang Avenue, Singapore 639798, 25 July 2007.
[2] Murray, M. P., “Gait as a total pattern of movement”, American journal
of Physical medicine, 46(1):290-333, 1967. Shanawaz Ahmed J received his MCA
[3] Davrondzhon Gafurov ,Einar Snekkenes ,Patrick Bours , “Improved Gait degree from the Department of Computer
Recognition Performance Using Cycle Matching”, In proceedings of Science Bharathidasan University, India. After that he
IEEE, 24th International Conference on Advanced Information
Networking and Applications Workshops, 2010. obtained the Masters of Philosophy degree from Periyar
[4] Aristotle (350BC), “On the Gait of Animals”, Translated by A. S. L. University, India. He is presently pursuing his PhD degree
Farquharson 2007. in Anna University, India. During 2004-2007, he was a
[5] Sarkar, S. , Phillips, P. J., Liu, Z.,Vega I. R., Grother, P., and Bowyer, lecturer in the Department of Computer Science The New
K., “The humanID gait challenge problem: Data sets,performance and College, Chennai, India. Since 2007, he has been serving as
analysis”, IEEE Trans.Pattern Anal. Mach. Intell., vol. 27, no. 2,pp.
162–177, Feb. 2005.
a Lecturer in college of computer science at King Khalid
[6] Liang Wang, Tieniu Tan, Huazhong Ning, and Weiming Hu, “Silhouette
University, Abha, KSA. His research interests include
Analysis-Based Gait Recognition for Human Identification”, IEEE image processing and image retrieval.
Transactions on pattern analysis and machine intelligence, vol. 25, no.
12, december 2003. Md. Ekramul Hamid received his B.Sc
[7] James B. Hayfron-Acquah, Mark S. Nixon, John N. Carter, ”Automatic and M.Sc degree from the Department of
gait recognition by symmetry analysis”, Image, Speech and Intelligent
Systems Group, Department of Electronics and Computer Science, Applied Physics and Electronics, Rajshahi
University of Southampton, Southampton, S017 1BJ, United Kingdom. University, Bangladesh. After that he
[8] Dawson, M. R., ”Gait Recognition”, Imperial College of Science, obtained the Masters of Computer Science
Technology & Medicine, London, June, 2002. from Pune University, India. He received his PhD degree
[9] Han, X.,”Gait Recognition Considering Walking Direction”, University from Shizuoka University, Japan. During 1997-2000, he
of Rochester, USA, August 20, 2010. was a lecturer in the Department of Computer Science and
[10] Haritaoglu, I., Harwood, D., Davis, L.”A real-time system for detecting Engineering, Rajshahi University. Since 2007, he has been
and tracking people in 2.5D”, Proceedings of the 5th European Conf.
Computer Vision 1998, Freiburg Germany, 1, pp.877-892 ,1998. serving as an associate professor in the same department.
[11] Amos Y. Johnson and Aaron F. Bobick. “A Multi-view Method for Gait He is currently working as an assistant professor in the
Recognition Using Static Body Parameters”.Electrical and Computer college of computer science at King Khalid University,
Engineering Georgia Tech., Atlanta, GA 30332. Abha, KSA. His research interests include Digital Signal
Processing and Speech Enhancement.
AUTHORS PROFILE
Dr. R.S.D Wahidabanu received her BE (Electronics &
Communication) and ME degree (Applied Electronics)
Mohamed Rafi received his BE and ME from University of Madras Chennai, India. Obtained PhD
degree in Computer Science & Engineering from Anna University, Tamil nadu, India. Having more
from Bangalore University, India. Presently than 30 years of experience in Teaching and research.
Pursuing PhD from Vinayaka Mission Working as Professor & Head, Dept of Electronics &
University, Salem, Tamil nadu, India. From communication engineering, Government college of
August 2007 to till date working as a Professor, Dept of Engineering, Salem. More than 13 students obtained phd
Computer Science & Engineering, HMS institute of and more than 20 students are pursuing phd under the
Technology, Tumkur, Karnataka, India. From November guidance. Published more than 30 papers in international
2001 to July 2007 Worked as Assistant Professor in the journals.
Department of Computer Science and Information
Technology, at Adama University, Ministry of Education,
Ethiopia. His research interests include Image Processing,
Database system and software engineering.

94 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Mining Fuzzy Cyclic Patterns
F. A. Mazarbhuiya M. A. Khaleel P. R. Khan
Department of Computer Science Department of Computer Science Department of Computer Science
College of Computer Science College of Computer Science College of Computer Science
King Khalid University King Khalid University King Khalid University
Abha, Kingdom of Saudi Arabia Abha, Kingdom of Saudi Arabia Abha, Kingdom of Saudi Arabia
e-mail: fokrul_2005@yahoo.com e-mail: khaleel_dm@yahoo.com e-mail: drpervaizrkhan@yahoo.com

Abstract— The problem of mining temporal association rules to define the similarity between fuzzy time intervals associated
from temporal dataset is to find association between items that with a frequent itemsets. Similarly, fuzzy distance between any
hold within certain time intervals but not throughout the dataset. two consecutive fuzzy time intervals associated with a frequent
This involves finding frequent sets that are frequent at certain itemset can be used to find the fuzzy time gaps between the
time intervals and then association rules among the items present fuzzy time intervals and its variance can be used to define
in the frequent sets. In fuzzy temporal datasets as the time of similarity among fuzzy time gaps associated with the frequent
transaction is imprecise, we may find set of items that are itemsets.
frequent in certain fuzzy time intervals. We call these as fuzzy
locally frequent sets and the corresponding associated association In section II we give a brief discussion on the works related
rules as fuzzy local association rules. The algorithm discussed [5] to our work. In section III we describe the definitions, terms
finds all fuzzy locally frequent itemsets where each frequent item and notations used in this paper. In section IV, we give the
set is associated with a list of fuzzy time intervals where it is algorithm proposed in this paper for mining fuzzy locally
frequent. The list of fuzzy time intervals may exhibit some frequent sets. We conclude with conclusion and lines for future
interesting properties e.g. the itemsets may be cyclic in nature. In work in section V. In the last section we give some references.
this paper we propose a method of finding such cyclic frequent
itemsets. II. RELATED WORKS
Keywords- Fuzzy time-stamp, Fuzzy time interval, Fuzzy The problem of discovery of association rules was first
temporal datasets, Fuzzy locally frequent sets, Fuzzy distance, formulated by Agrawal et al in 1993. Given a set I, of items
Variance of a fuzzy interval and a large collection D of transactions involving the items,
the problem is to find relationships among the items i.e. the
I. INTRODUCTION presence of various items in the transactions. A transaction
The problem of mining association rules has been defined t is said to support an item if that item is present in t. A
initially [10] by R. Agarwal et al for application in large super transaction t is said to support an itemset if t supports each
markets. Large supermarkets have large collection of records of of the items present in the itemset. An association rule is
daily sales. Analyzing the buying patterns of the buyers will an expression of the form X ⇒ Y where X and Y are subsets
help in taking typical business decisions such as what to put on of the itemset I. The rule holds with confidence τ if τ% of
sale, how to put the materials on the shelves, how to plan for the transaction in D that supports X also supports Y. The
future purchase etc.
rule has support σ if σ% of the transactions supports X ∪ Y.
Mining for association rules between items in temporal A method for the discovery of association rules was given
databases has been described as an important data-mining in [9], which is known as the A priori algorithm. This was
problem. Transaction data are normally temporal. The market then followed by subsequent refinements, generalizations,
basket transaction is an example of this type. extensions and improvements.
Mining fuzzy temporal dataset is also an interesting data Temporal Data Mining is now an important extension
mining problem. In [5], author proposed a method of finding of conventional data mining and has recently been able to
fuzzy locally frequent sets from such datasets. The algorithm attract more people to work in this area. By taking into
discussed in [5] extracts all frequent itemsets along with a set account the time aspect, more interesting patterns that are
of list of fuzzy time intervals where each frequent itemset is time dependent can be extracted. There are mainly two
associated a list of fuzzy time intervals. The list of fuzzy time broad directions of temporal data mining [6]. One concerns
intervals associated with a frequent itemsets can be used to find the discovery of causal relationships among temporally
some interesting results. In this paper we propose to study the oriented events. Ordered events form sequences and the
problem of cyclic nature of a list of time intervals associated cause of an event always occur before it. The other
with a frequent itemset and devise a method to extract all concerns the discovery of similar patterns within the same
frequent itemsets which are cyclic. We call such frequent time sequence or among different time sequences. The
itemset as fuzzy cyclic frequent itemsets as the time intervals underlying problem is to find frequent sequential patterns
are fuzzy in nature. In such case, as the variance of the fuzzy in the temporal databases. The name sequence mining is
intervals is invariant with respect to translation, it can be used normally used for the underlying problem. In [7] the

95 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

problem of recognizing frequent episodes in an event In general, a generalized fuzzy number A is described as
sequence is discussed where an episode is defined as a any fuzzy subset of the real line R, whose membership function
collection of events that occur during time intervals of a A(x) satisfying the following conditions
specific size. (1) A(x) is continuous mapping from R to the closed
The association rule discovery process is also extended interval [0, 1]
to incorporate temporal aspects. In temporal association
rules each rule has associated with it a time interval in (2) A(x)=0, -∝< x ≤ c
which the rule holds. The problems associated are to find (3) A(x)= L(x) is strictly increasing on [c, a]
valid time periods during which association rules hold, the
discovery of possible periodicities that association rules (4) A(x)=w, a ≤ x ≤ b
have and the discovery of association rules with temporal (5) A(x)= R(x) is strictly decreasing on [b, d]
features. In [8], [14], [15] and [16], the problem of
temporal data mining is addressed and techniques and (6) A(x)=0, d ≤ x<∝,
algorithms have been developed for this. In [8] an where 0 <w ≤1, a, b, c and d are real numbers. This type of
algorithm for the discovery of temporal association rules is generalized fuzzy number is denoted by A=(c, a, b, d; w)LR.
described. In [2], two algorithms are proposed for the When w=1, the above generalized fuzzy number will be a
discovery of temporal rules that display regular cyclic fuzzy interval and is denoted by A=(c, a, b, d)LR. When L(x)
variations where the time interval is specified by user to and R(x) are straight line, then A is a trapezoidal fuzzy number
divide the data into disjoint segments like months, weeks, and is denoted by (c, a, b, d). If a=b, then the above trapezoidal
days etc. Similar works were done in [4] and [17] number will be a triangular fuzzy number denoted by (c, a, d).
incorporating multiple granularities of time intervals (e.g.
The h-level of the fuzzy number [t1-a, t1, t1+a] is a closed
first working day of every month) from which both cyclic
interval [t1+(α-1).a, t1+(1-α).a]. Similarly the h-level of the
and user defined calendar patterns can be achieved. In [1],
fuzzy interval [t1-a, t1, t2, t2+a] is a closed interval [t1+(α-1).a,
the method of finding locally and periodically frequent sets
t2+(1-α).a].
and periodic association rules are discussed which is an
improvement of other methods in the sense that it Chen and Hsieh [11, 12, 13] proposed graded mean integration
dynamically extracts all the rules along with the intervals representation for generalized fuzzy number as follows:
where the rules hold. In ([18], [19]) fuzzy calendric data Suppose L-1 and R-1are inverse functions of the functions L and
mining and fuzzy temporal data mining is discussed where R respectively and the graded mean h-level value of
user specified ill-defined fuzzy temporal and calendric generalized fuzzy number A=(c, a, b, d; w)LR is h[L-1(h)+ R-
1
patterns are extracted from temporal data. (h)]/2. Then the graded mean integration representation of
In [5], the authors propose a method of extracting generalized fuzzy number based on the integral value of graded
fuzzy locally frequent sets from fuzzy temporal datasets. mean h-levels is
The algorithm discussed in [5], extracts all frequent w w
itemsets with each itemset is associated with a list fuzzy L−1 ( h ) + R −1 ( h )
time intervals where the itemset is frequent. The list fuzzy
time intervals associated with a frequent itemset exhibits
P(A)= ∫
0
h( 2 )dh / ∫ hdh
0
some interesting properties e.g. the size of the fuzzy time
intervals may be almost same and also the time gap where h is between 0 and w, 0<w≤1.
between two consecutive fuzzy time intervals is also almost B. Fuzzy distance
same. We call such frequent itemset as a fuzzy cyclic
frequent itemset. So the study is basically an intra-itemset Chen and Wang [13] proposed fuzzy distance between any two
study. trapezoidal fuzzy numbers as follows: Let A=(a1, a2, a3, a4),
B=(b1, b2, b3, b4) be two trapezoidal fuzzy numbers, and their
III. PROBLEM DEFINITION graded mean integration representation are P(A), P(B)
respectively. Assume
A. Some Definition related to Fuzzy sets
si=(ai-P(A)+bi-P(B))/2, i=1, 2, 3, 4;
Let E be the universe of discourse. A fuzzy set A in E is
characterized by a membership function A(x) lying in [0,1]. ci=P(A)-P(B)+ si , i=1, 2, 3, 4;
A(x) for x ∈ E represents the grade of membership of x in A.
Thus a fuzzy set A is defined as then the fuzzy distance of A, B is C=(c1, c2, c3, c4). Obviously
the fuzzy distance between two trapezoidal numbers is also a
A={(x, A(x)), x ∈ E } trapezoidal number.
A fuzzy set A is said to be normal if A(x) =1 for at least one C. Possibilistic variance of a fuzzy number
x∈E Let F be a family of fuzzy number and A be a fuzzy number
belonging to F. Let Aγ =[a1(γ), a2(γ)], γ∈[0, 1] be a γ-level of

96 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

A. Carlsson and Fuller [3] defined the possibilistic variance of gaps) and see whether it is almost equal to the variance of the
fuzzy number A∈ F as fuzzy distance (time gap) between the third and the fourth
fuzzy time intervals. If the average of the variance of the first
1 two fuzzy time intervals is almost equal with the variance of
Var(A) = 1
2 ∫ γ (a 2 (γ ) − a1 (γ )) 2 dγ the third interval we proceed further or otherwise stop. In
0
general if the average of the variance of the first (n-1) fuzzy
Before proceeding further we review the theorem given by time intervals is almost equal to the variance of the n-th fuzzy
Carlsson and Fuller [3]. time interval and the average of variance of first (n-2) fuzzy
distances (time gaps) are almost equal to the n-1 th time gap,
D. Theorem: The variance of a fuzzy number is invariant to then the average of variances of n fuzzy time intervals is
shifting. compared with variance of (n+1)th fuzzy time interval and that
of the first n-1 time gaps is compared with the n th time gap.
Proof: Let A∈ F be a fuzzy number and let θ be a real number. This way we can extract fuzzy cyclic patterns if such patterns
If A is shifted by value θ, then we get a fuzzy number, denoted exist. We describe below the algorithm for extracting periodic
by B, satisfying the property B(x)=A(x-θ) for all x∈ R. Then item sets.
from the relationship
Algorithm for extracting fuzzy cyclic frequent item sets
Bγ=[a1(γ)+θ, a2(γ)+θ] For each fuzzy locally frequent item set iset do
we find { t1 first fuzzy time interval for iset
1
v1 var(t1)
Var(B)= 1
2 ∫ γ ((a 2 (γ ) + θ ) − (a1 (γ ) + θ )) 2 dγ
0
t2 second fuzzy time interval for iset

1
v2 var( t2)
=1
2 ∫ γ (a 2 (γ ) − a1 (γ )) dγ
2
if not almostequal(v1, v2,sgma) then
0
{ report that iset is not fuzzy cyclic in nature
=Var(A) continue /* go for the next frequent item set */
E. Almost equal fuzzy intervals }
Given two fuzzy intervals A and B, we say A is almost equal B n=1
or B is almost equal to A if the variances of both A and B are
equal up to a small variation say λ%. i.e. ftg1 fuzzydist(t1,t2)
var(A)+ λ% of var(B)=var(B) v(ftg1) var(fuzzydist(t1,t2))
var(B)+ λ% of var(A)=var(A) avgvar (v1 +v2)/2
where λ is specified by the user. flag = 0

IV. PROPOSED ALGORITHM while not end of fuzzy interval list for iset do
A. Extraction of fuzzy cyclic patterns { tint current fuzzy time interval
One way to extract these sets is to find the fuzzy distance ftg fuzzydist(tint, t2)
between any two consecutive fuzzy time intervals of the same v(ftg) var(fuzzydist(tint, t2))
frequent set. If the fuzzy distance (time gap) between any two
consecutive frequent time intervals are found to be almost if almostequal(v(ftg),v(ftg1), sgma) then
equal and also the fuzzy time intervals are found to be almost avv(gftg) (n*v(ftg1) +v(ftg))/(n+1)
equal (the definition of almost equal fuzzy time interval is
given in Definition E of section III) then we call these frequent else
sets as fuzzy cyclic frequent sets. Now to find out such type of
{ flag = 1; break;}
cyclic nature for each frequent item set we proceed as follows.
If the first fuzzy time interval is almost equal to the second var var(tint)
fuzzy time interval then we see whether the fuzzy distance
if almostequal(var, avgvar, sgma) then
(time gap) between the first and the second fuzzy time interval
is almost equal to the fuzzy distance (time gap) between the avgvar ((n+1)*avgvar +var) /(n+2)
second and third fuzzy time intervals. If it is, then we take the
else
average of the variance of the first two fuzzy distances (time

97 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

{flag = 1; break;} [3] C. Carlsson and R. Fuller; On Possibilistic Mean Value and Variance of
Fuzzy Numbers, Fuzzy Sets and Systems 122 (2001), pp. 315-326.
t2 tint; [4] G. Zimbrao, J. Moreira de Souza, V. Teixeira de Almeida and W. Araujo
da Silva; An Algorithm to Discover Calendar-based Temporal
n n+1; Association Rules with Item’s Lifespan Restriction, Proc. of the 8th
ACM SIGKDD Int’l Conf. on Knowledge Discovery and Data Mining
} (2002) Canada, 2nd Workshop on Temporal Data Mining, v. 8, 701-706,
2002.
if (flag == 1)
[5] F. A. Mazarbhuiya, and Ekramul Hamid; Finding Fuzzy Locally
report that iset is not fuzzy cyclic in nature Frequent itemsets, International Journal of Computer Science and
Information Security, ISSN: 1947-5500, LJS Publisher and IJCSIS Press,
else USA, 2011
[6] J. F. Roddick, M. Spillopoulou; A Biblography of Temporal, Spatial and
report that iset is fuzzy cyclic in nature Spatio-Temporal Data Mining Research; ACM SIGKDD, June 1999.
[7] H. Manilla, H. Toivonen and I. Verkamo; Discovering frequent episodes
} in sequences; KDD’95; AAAI, 210-215, August 1995.
Here the function var() returns the fuzzy variance value of the [8] J. M. Ale and G.H. Rossi; An approach to discovering temporal
association rules; Proceedings of the 2000 ACM symposium on Applied
fuzzy time intervals associated with a fuzzy locally frequent Computing, March 2000.
set, fuzzydist() returns fuzzy distance between two consecutive [9] R. Agrawal and R. Srikant; Fast algorithms for mining association rules,
fuzzy time intervals associated with a fuzzy locally frequent Proceedings of the 20th International Conference on Very Large
set i.e. fuzzy time gap, v() returns fuzzy variance value of the Databases (VLDB ’94), Santiago, Chile, June 1994.
fuzzy time gap between two consecutive fuzzy time intervals [10] R. Agrawal, T. Imielinski and A. Swami; Mining association rules
associated with a fuzzy locally frequent itemset, avv() returns between sets of items in large databases; Proceedings of the ACM
SIGMOD ’93, Washington, USA, May 1993.
the average of the variances of fuzzy time gaps avgvar
[11] S. H. Chen and C. H. Hsieh; Graded mean integration representation of
returns average of the fuzzy variances of fuzzy time intervals generalised fuzzy number, Proc. of Conference of Taiwan Fuzzy System
associated with a fuzzy locally frequent set, sgma is the Association , Taiwan, 1998.
threshold value upto which the variation of fuzzy variance [12] S. H. Chen and C. H. Hsieh; Graded mean integration representation of
values can be permitted. generalised fuzzy number, Journal of the Chinese Fuzzy Sytem
Association, Vol. 5, No. 2, pp. 1-7, Taiwan, 1999.
V. CONCLUSIONS AND LINES FOR FUTURE WORKS [13] S. H. Chen and C. H. Hsieh; Representation, Ranking, Distance,
Similarity of L-R type fuzzy number and Application, Australia Journal
An algorithm for finding fuzzy cyclic frequent sets from fuzzy of Intelligent Information Processing Systems, Vol. 6, No. 4, pp. 217-
temporal data is given in the paper. The algorithm [5] gives all 229, Australia.
fuzzy locally frequent itemsets where each frequent itemsets is [14] X. Chen and I. Petrounias; A framework for Temporal Data Mining;
associated with a list of fuzzy time intervals where it is Proceedings of the 9th International Conference on Databases and Expert
frequent. Our algorithm takes the input from the result of the Systems Applications, DEXA ’98, Vienna, Austria. Springer-Verlag,
Berlin; Lecture Notes in Computer Science 1460, 796-805, 1998.
algorithm [5]. As the the variance of the fuzzy interval is
[15] X. Chen and I. Petrounias; Language support for Temporal Data Mining;
invariant with respect to shifting, we can consider two fuzzy Proceedings of 2nd European Symposium on Principles of Data Mining
intervals having same variance value belonging in two different and Knowledge Discovery, PKDD ’98, Springer Verlag, Berlin, 282-
parts of time domain as equal. Using this notion we define the 290, 1998.
equality of the fuzzy time intervals associated with a fuzzy [16] X. Chen, I. Petrounias and H. Healthfield; Discovering temporal
locally frequent itemsets. Our algorithm supplies all frequent Association rules in temporal databases; Proceedings of IADT’98
sets which are fuzzy cyclic. (International Workshop on Issues and Applications of Database
Technology, 312-319, 1998.
We may have some patterns where the fuzzy time gaps are [17] Y. Li, P. Ning, X. S. Wang and S. Jajodia; Discovering Calendar-based
almost equal but the fuzzy time intervals of frequency are not Temporal Association Rules, In Proc. of the 8th Int’l Symposium on
equal even in the approximate sense. We may also have some Temporal Representation and Reasonong, 2001.
patterns where the fuzzy time-gaps are not equal but durations [18] W. J. Lee and S. J. Lee; Discovery of Fuzzy Temporal Association
Rules, IEEE Transactions on Systems, Man and Cybenetics-part B;
of the time intervals are almost equal. The above algorithm can Cybernetics, Vol 34, No. 6, 2330-2341, Dec 2004.
be modified accordingly to find such patterns. In future we [19] W. J. Lee and S. J. Lee; Fuzzy Calendar Algebra and It’s Applications to
would like to devise methods to find various kinds of patterns Data Mining, Proceedings of the 11th International Symposium on
which may exist in fuzzy temporal datasets. Temporal Representation and Reasoning (TIME’04), IEEE, 2004.
References AUTHOR’S PROFILE
[1] A. K. Mahanta, F. A. Mazarbhuiya and H. K. Baruah; Finding Locally Fokrul Alom Mazarbhuiya received B.Sc. degree in Mathematics
and Periodically Frequent Sets and Periodic Association Rules, from Assam University, India and M.Sc. degree in Mathematics from
Proceeding of 1st Int’l Conf on Pattern Recognition and Machine
Intelligence (PreMI’05),LNCS 3776, 576-582, 2005.
Aligarh Muslim University, India. After this he obtained his Ph.D.
degree in Computer Science from Gauhati University, India. Since
[2] B. Ozden, S. Ramaswamy and A. Silberschatz; Cyclic Association
Rules, Proc. of the 14th Int’l Conference on Data Engineering, USA,
2008 he has been serving as an Assistant Professor in College of
412-421, 1998. Computer Science, King Khalid University, Abha, kingdom of Saudi
Arabia. His research interest includes Data Mining, Information
security, Fuzzy Mathematics and Fuzzy logic.

98 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Mohammed Abdul Khaleel received B.Sc. degree in Mathematics
from Osmania University, India and M.C.A degree from Osmania
University, India. After that worked in Global Suhaimi Company
Dammam Saudi Arabia as Senior Software Developer.Since 2008
serving as Lecturer at College of Computer Science, King Khalid
University, Abha, Kingdom of Saudi Arabia. His research interest
includes Data Mining, Software Engineering.
Pervaiz Rabbani Khan received his B.Sc. B.Sc.(Hons) and M.Sc. in
Physics from Punjab University, Lahore, Pakistant. After this he has
done M.Sc. in Computer Science from New Castle upon Tyne, U.K.
After this he obtained his Ph.D. from the same University. Since
2001 he has been working as an Assistant Professor in College of
Computer Science, King Khalid University, Abha, Kingdom of Saudi
Arabia. His research interest includes Simulation and Modeling,
Fuzzy Logic.

99 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Robust Color Image Watermarking Using
Nonsubsampled Contourlet Transform
C.Venkata Narasimhulu K.Satya Prasad
Professor, Dept of ECE Professor, Dept of ECE,
HIET, Hyderabad, India JNTU Kakinada, India
narasimhulucv@gmail.com prasad_kodati@yahoo.co.in,

Abstract- applications, embedded watermark should be invisible,
In this paper, we propose a novel hybrid spread robust and have a high capacity. Invisibility refers to
spectrum watermarking scheme for authentication of degree of distortion introduced by the watermark and its
color images using nonsubsampled contourlet transform affect on the viewers and listeners. Robustness is the
and singular value decomposition. The host color image resistance of an embedded watermark against
and color watermark images are decomposed into intentional attack and normal signal processing
directional sub- bands using Nonsubsampled contourlet operations such as noise, filtering, rotation, scaling,
transform and then applied Singular value decomposition cropping and lossey compression etc. Capacity is the
to mid frequency sub-band coefficients. The singular amount of data can be represented by embedded
values of mid frequency sub-band coefficients of color
watermark.[1]
watermark image are embedded into singular values of
mid frequency sub-band coefficients of host color image in Watermarking techniques may be classified in
Red, Green and Blue color spaces simultaneously based on different ways. The classification may be based on the
spread spectrum technique. The experimental results type of watermark being used, i.e., the watermark may
shows that the proposed hybrid watermarking scheme is
robust against common image processing operations such
be a visually recognizable logo or sequence of random
as, JPEG, JPEG 2000 compression, cropping, Rotation, numbers. A second classification is based on whether
histogram equalization, low pass filtering ,median the watermark is applied in the spatial domain or the
filtering, sharpening, shearing ,salt & Pepper noise, transform domain. In spatial domain, the simplest
Gaussian noise, grayscale conversion etc. It has also been method is based on embedding the watermark in the
shown the variation of visual quality of watermarked least significant bits (LSB) of image pixels. However,
image for different scaling factors. The comparative spatial domain techniques are not resistant enough to
analysis reveals that the proposed watermarking scheme image compression and other image processing
out performs the color image watermarking schemes operations.
reported recently.
Keywords: Color image watermarking, Nonsubsampled
Transform domain watermarking schemes such as
Contourlet Transform, Singular value decomposition, Peak those based on the discrete cosine transform (DCT), the
signal to noise ratio, normalized Correlation coefficient. discrete wavelet transform (DWT), contourlet
transforms along with numerical transformations such
1. INTRODUCTION: as Singular value Decomposition (SVD) and Principle
In recent years, multimedia products were rapidly component analysis (PCA) typically provide higher
distributed over the fast communication systems such image fidelity and are much robust to image
as Internet, so there exist strong requirement to protect manipulations.[2]Of the so far proposed algorithms,
the ownership and authentication of the multimedia wavelet domain algorithms perform better than other
data. Digital watermarking is a method of securing the transform domain algorithms since DWT has a number
digital data by embedding additional information called of advantages over other transforms including time
water mark into the digital multimedia content. This frequency localization, multi resolution representation,
embedding information can be later extracted from or superior HVS modeling, and linear complexity and
detected in the multimedia to make an assertion about adaptively and it has been proved that wavelets are
the data authenticity. Digital watermarks remain intact good at representing point wise discontinuities in one
under transmission/transformation, allowing us to dimensional signal. However, in higher dimensions,
protect our ownership rights in digital form. Absence of e.g. image, there exists line or curve-shaped
watermark in a previously watermarked image would discontinuities. Since, 2D wavelets are produced by
lead to the conclusion that the data content has been tensor products of 1D wavelets; they can only identify
modified. A watermarking algorithm consists of horizontal, vertical, diagonal discontinuities (edges) in
watermark structure, an embedding algorithm and images, ignoring smoothness along contours and
extraction or detection algorithm. In multimedia curves. Curvelet transform was defined to represent two

100 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
dimensional discontinuities more efficiently, with least 2. NONSUBSAMPLED CONTOURLET
square error in a fixed term approximation. Curvelet TRANSFORM
transform was proposed in continuous domain and its
discretisation was a challenge when critical sampling is The Nonsubsampled contourlet transform is a new
desired. Contourlet transform was then proposed by DO image decomposition scheme introduced by Arthur
and Vetterli as an improvement of Curvelet transform. L.Cunha, Jianping Zhou and Minh N.Do [8]. NSCT is
The Contourlet transform is a directional multi more effective in representing smooth contours in
resolution expansion which can represents images different directions of in an image than contourlet
contains contours efficiently. The CT employs transform and discrete wavelet transform. The NSCT is
Laplacian pyramids to achieve multi resolution fully shift invariant, Multi scale and multi direction
decomposition and directional filter banks to achieve expansion that has a fast implementation. The NSCT
directional decomposition [3] Due to down sampling exhibits a similar sub band decomposition as that of
and up sampling, the Contourlet transform is Shift contourlets, but without down samplers and up samplers
variant. However shift invariance is desirable in image in it. Because of its redundancy the filter design problem
analysis applications such as edge detection, Contour of nonsubsampled contourlet is much less constrained
characterization, image enhancement [4] and image than that of contourlet. The NSCT is constructed by
watermarking. Here, we present a NonSubsampled combining nonsubsampled pyramids and
Contourlet transform (NSCT) [5] which is shift nonsubsampled directional filter bank as shown in
invariant version of the contourlet transform. The figure (1).The nonsubsampled pyramid structure results
NSCT is built upon iterated nonsubsampled filter banks the multi scale property and nonsubsampled directional
to obtain a shift invariant image representation.
filter bank results the directional property.
In all above transform domain watermarking techniques
including NSCT the watermarking bits would be
directly embedded in the locations of sub band
coefficients. Though here the visual of perception of
original image is preserved, the watermarked image
when subjected to some intentional attacks like
compression the watermark bits will get damaged.
Coming to the spatial domain watermarking using
numerical transformation like SVD (Gorodetski [6], liu
et al [7]) they provide good security against tampering
and common manipulations for protecting rightful
ownership. But these schemes are non adaptive, thus
unable to offer consistent perceptual transparency of
watermarking of different images. To provide adaptive
transparency, robustness to the compressions and (a) (b)
insensitivity to malicious manipulations, we propose a Figure 1 The nonsubsampled contourlet transform (a)
novel image hybrid watermarking scheme using NSCT nonsubsampled filter bank structure that implements the NSCT.
and SVD. (b) Idealized frequency partitioning obtained with NSCT

2.1 Nonsubsampled Pyramids
In this paper, proposed method is compared with
another which is based on Contourlet Transform and
singular value decomposition (CT-SVD). The peak The nonsubsampled pyramid is a two channel
signal to noise ratio (PSNR) between the original image nonsubsampled filter bank as shown in figure
and watermarked image and the normalized correlation 2(a).The H0(z) is the low pass filter and one then sets
coefficients (NCC) and bit error rate (BER) between H1(z) =1-H0(z). the corresponding synthesis filters
the original watermark and extracted were calculated G0(z) =G1(z)=1.
with and without attacks. The results show high
improvement detection reliability using proposed the perfect reconstruction condition is given by
method. The rest of this paper is organized as follows. Bezout identity
Section 2 describes the Nonsubsampled contourlet
transform, section 3 describes singular value
decomposition, section 4 illustrates the details of H0(z)G0(z)+H1(Z) G1 (Z) =1………………(1)
proposed method, in section 5 experimental results are
discussed without and with attacks, conclusion and
future scope are given in section 6.

101 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
:

(a) (b)
Figure (2): Nonsubsampled pyramidal filter (a). Ideal frequency response of nonsubsampled pyramidal filter
(b).The cascading analysis of three stages nonsubsampled pyramid by iteration of two channels
Nonsubsampled filter banks .

Multi scale decomposition is achieved from invariant is constructed by eliminating the down and
nonsubsampled pyramids by iterating the up samplers in the DFB.The ideal frequency response
nonsubsampled filter banks by up sampling all filters of nonsubsampled filter banks is shown in figure3 (a)
by 2 in both direction the next level decomposition is
achieved. The complexity of filtering is constant To obtain multi directional decomposition, the
whether the filtering is with H(z) or an up sampled nonsubsampled DFBs are iterated. To obtain the
filter H(z m ) computed a Trous algorithm The next level decomposition, all filters are up
cascading of three stage analysis part is shown in sampled by a quincunx matrix given by
figure 2( b)
1 1
2.2 Nonsubsampled directional Filter Banks: Q=
The directional filter bank (DFB) is constructed from 1 ‐1 ……………..(2)
the combination of critically-sampled two-channel
fan filter banks and resampling operations. The
outcome of this DFB is a tree-structured filter bank
splitting the 2-D frequency plane into wedges. The The analysis part of iterated nonsubsampled filter
nonsubsampled directional filter bank which is shift bank is shown in figure 3 (b)

(a) (b)
Figure (3) Nonsubsampled directional filter bank (a) idealized frequency response of nonsubsampled directional filter bank.(b) The
analysis part of an iterated nonsubsampled directional bank.

3. SINGULAR VALUE DECOMPOSITION data, for data compression and data denoising. If
A is any N x N matrix, it is possible to find a
Singular value decomposition (SVD) is a decomposition of the form
popular technique in linear algebra and it has
applications in matrix inversion, obtaining low
dimensional representation for high dimensional

3
102 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
A=USVT including discrete cosine transform (DCT), discrete
wavelet transform (DWT), Contourlet transform
(CT) etc have been used to embed watermark into
original image. here the proposed scheme uses
nonsubsampled contourlet transform(NSCT) along
with SVD for watermarking to obtain better
performance compared to existing hybrid
algorithms.
4. PROPOSED ALGORITHM
Where U and V are orthogonal matrices of order
N x N and N x N such that UTU=I,VTV=I , and the In this paper, Nonsubsampled Contourlet
diagonal matrix S of order N x N has elements λi Transform and SVD based hybrid technique is
(i=1,2,3,..n) , I is an identity matrix of order N x N. proposed for color image watermarking that uses
The diagonal entries are called singular values of true color images for both watermark and host
matrix A, the columns of U matrix are called the left images. The robustness and visual quality of
singular values of A, and the columns of V are watermarked image is tested with three quantifiers
called as the right singular values of A. such as PSNR, NCC and Bit Error Rate. It is
The general properties of SVD are [1], [2], [9] investigated whether the NSCT-SVD advantages
a) Transpose: A and its transpose AT have the over CT-SVD for color image watermarking with
same non-zero singular values. their extra features would provide any significance
b) Flip: A, row-flipped Arf, and column- in terms of watermark robustness and invisibility.
flipped Acf have the same non-zero singular values. 4.1 , 4.2 explain the watermark embedding and
c) Rotation: A and Ar (A rotated by an extraction algorithm [10],[11]
arbitrary degree) have the same non-zero singular 4.1 Watermark Embedding Algorithm
values.
d) Scaling: B is a row-scaled version of A by The proposed watermark embedding algorithm
repeating every row for L1 times. For each non-zero is shown in Figure 4. The steps of watermark
singular value λ of A, B has √L1λ. C is a column- embedding algorithm are as follows.
scaled version of A by repeating every column for Step1: Separate the R G B color spaces of both
L2 times. For each nonzero singular value λ of A, C host and watermark color images.
has √L2λ. If D is row-scaled by L1 times and
column-scaled by L2 times, for each non-zero Step2: Apply Nonsubsampled Contourlet
singular value λ of A, D has √L1L2λ. Transform to the R color space of both host image
e) Translation: A is expanded by adding rows and watermark image to decompose them into sub
and columns of black pixels. The resulting matrix bands.
Ae has the same Non-zero singular values as A. Step3: Apply SVD to mid frequency sub-band of
The important properties of SVD from the view CT of R color space of both host and watermark
point of image processing applications are: image.
1. The singular values of an image have very Step4: Modify the singular values of mid
good stability i.e. When a small perturbation is frequency sub-band coefficients of R color space of
added to an image, their singular values do not host image with the singular values of mid
change significantly. frequency sub-band coefficients of R color space of
watermark image using spread spectrum technique.
2. Singular value represents intrinsic algebraic
image properties. i.e. λI’ = λI + α λW.,
Due to these properties of SVD, in the last few Where α is scaling factor [9], λI is singular value
years several watermarking algorithms have been of R color space of host image, λW is singular value
proposed based on this technique. The main idea of of R color space of watermark and λI’ becomes
this approach is to find the SVD of a original image singular value of R color space watermarked image.
and then modify its singular values to embedded the Step5: Apply inverse SVD on modified singular
watermark. Some SVD based algorithms are purely values obtained in step4 to get the mid frequency
SVD based in a sense that only SVD domain is used sub-band coefficients of watermarked image.
to embed watermark into original image. Recently
some hybrid SVD based algorithms have been Step6: Apply inverse Nonsubsampled
proposed where different types of transform domain Contourlet Transform to the mid frequency sub-

4
103 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
band coefficients obtained in step 5 to get the R Step5: Apply inverse SVD to obtain mid
color space of watermarked image. frequency coefficients of R color space of
transformed watermark image using Step 3.
Step7: Apply the same Steps from Step2 to
Step6 for the G and B color subspaces. Step6: Apply inverse NSCT using the
coefficients of the mid frequency sub-band to obtain
Step 8: Combine the R,G and B color spaces of
the R color space of Watermark image.
watermarked image to obtain the color watermarked
image. Step7: Repeat the Steps 2 to 6 for G and B color
spaces.
Step8: Combine the R,G and B color spaces to
get the color watermark.

Figure 4 Watermark Embeddign Algorithm
Figure 5 Watermark Extracting Algorithm
4.2 Watermark Extraction Algorithm
The watermark extraction algorithm is shown in
Figure 5. The Steps of watermark extraction 5. EXPERIMENTAL RESULTS
algorithm are as follows. In the experiments, we use the true color
“tajmahal.jpg” of size 256X256 as host image as
Step1: Separate the R,G,B color spaces of
shown in the Figure 6 and true color “lena.jpg” of
watermarked image. size 128 X 128 as watermark as shown in Figure 7.
Step2: Apply Nonsubsampled Contourlet The experiment is performed by taking scaling
Transform to the R color space obtained in step1. factor alpha as 0.5.The results show that there are no
perceptibly visual degradations on the watermarked
Step3: Apply SVD to mid frequency sub-band of image shown in Figure 8 with a PSNR of
R color space of transformed watermarked image. 45.2253dB. Extracted watermark without attack is
shown in Figure 9 with NCC around unity and BER
Step4: Extract the singular values from mid
of 0.1339. MATLAB 7.6 version is used for testing
frequency sub-band of R color space of
the robustness of the proposed method.
watermarked and host image
The proposed algorithm is tested for different host
i, e λW = ( λI’ - λI )/ α images such as “lotus.jpg”, ”Baboon.jpg”,
Where λI is singular value of watermarked image. ”Barbara.jpg”, ”Way.jpg” ,”Horse.jpg” and
“Wheel.jpg” as shown in Table 1 and it is observed
that there are no visual degradations on the respected

5
104 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
watermarked images. For all the different Host test color space of host image using eq.3 [12]. The
images, the watermark is effectively extracted with final PSNR of watermarked image is taken as mean
around unity NCC. Various intentional and non- of PSNR obtained with three color spaces. The
intentional attacks are tested for robustness of the similarity of extracted watermark with original
proposed watermark algorithm includes watermark embedded is measured using NCC. The
JPEG,JPEG2000compressions, low pass filtering, NCC is calculated using eq. (4) [13]for the three
Rotation, Histogram Equalization ,Median Filtering, color spaces and their mean is taken as the resultant
Salt &Pepper Noise, Weiner Filtering, Gamma Normalized Correlation coefficient. The proposed
Correction, Gaussian Noise, Rescaling, Sharpening method is also tested for binary and grayscale
Blurring ,Contrast Adjustment ,Automatic cropping, watermark image of size 128x128 and watermarked
Dilation, Row Colum Copying, Row Colum and extracted watermark are shown in table 3.
removing, color to Gray scale conversion ,shearing
and sharpening. The term robustness describes the
watermark resistance to these attacks and can be
measured by the bit-error rate which, is defined as the
ratio of wrong extracted bits to the total number of ……….….(3)
embedded bits. Normalized Correlation Coefficient:
In table 2, extracted watermark and attacked
watermarked image with NCC and BER are shown.
The quality and imperceptibility of watermarked
image is measured by using PSNR. The PSNR is ………..(4)
calculated separately for R, G, B color space of
watermarked image with respect to the respective

Fig 6:Original image- Fig 7:Watermark image- Fig 8:Watermarked Lena Fig 9:Extracted
"Tajmahal.jpg” "Lena.jpg” PSNR= 45.2253 Watermark
Ncc=0.9991,Ber=0.1339

TABLE 1: WATERMARKED AND EXTRACTED WATERMARK WITH PSNR, NCC, AND BER FOR DIFFERENT ORIGINAL
IMAGES.

Original image Watermark image Watermarked image with Extracted image
“lotus.jpg” “LENA.jpg” PSNR=46.2785 NCC= 0.9983,Ber=0.1610

6
105 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Watermarked image with
Original image Watermark image Extracted image
PSNR=44.8322
“baboon.jpg” “LENA.jpg” NCC=0.9992, Ber=0.1342

Original image Watermark image Watermarked image with Extracted image
“barbara.jpg” “LENA.jpg” PSNR=44.4930 NCC=0.9994,Ber=0.1299

Original image Watermark image Watermarked image with Extracted image
“way.jpg” “LENA.jpg” PSNR=44.7550 NCC= 0.9994, Ber=0.1140

Original image Watermark image Watermarked image with Extracted image
“horse.jpg” “LENA.jpg” PSNR= 44.7308 NCC= 0.9994, Ber=0.1201

Original image Watermark image Watermarked image with Extracted image
“wheeljpg” “LENA.jpg” PSNR= 45.5204 NCC= 0.9985, Ber=0.1614

7
106 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

TABLE 2: EXTRACTED WATERMARKS WITH NCC AND BER FOR DIFFERENT ATTACKS ALONG WITH ATTACKED
WATERMARKED IMAGE

Jpeg compression Ncc= 0.9985,Ber=0.3306 Jpeg2000Ncc= 0.9995,Ber=0.1056

Salt & pepper noise Ncc= 0.6948, Ber=0.4503 Low Pass filtering Ncc= 0.9729 Ber=0.2995

utomatic cropping Ncc= 0.9538 Ber=0.3449 Histogram Equalization Ncc= 0.9808 Ber=0.3128

Rotation Ncc= 0. 0.9951 Ber=0.2958 Median filtering Ncc= 0.9484 Ber=0.3178

Contrast adjustment Ncc= 0.9985 Ber= 0.1613 Weiner filter Ncc= 0.9982 Ber=0.2051

8
107 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Gamma correction Ncc= 0.9989 Ber=0.1387 Gaussian Noise Ncc= 0.8399 Ber=0.3120

Sharpening Ncc= 0.8379 Ber=0.3967 Gaussian Blurring Ncc= 0.9719 Ber=0.3003

Shearing Ncc= 0.9744 Ber=0.2889 Dilatations= 0.9443 Ber=0.3332

Color to grayscale Ncc= 0.8163 Ber=0.3490 Row & column removal Ncc=0.9977 Ber=0.1930

9
108 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
Row column copying Ncc= 0.9902 Ber=0.9734 Scaling (150%) Ncc = 0.9187

TABLE 3: WATERMARKED AND EXTRACTED WATERMARK IMAGES FOR BINARY AND GRAYSCALE WATERMARK

Original image Binary Watermark image Watermarked image Extracted image
“tajmahal.jpg “ksp.bmp”. PSNR= 47.6710 Ncc= 0.9995, Ber=0.0157

Original image Binary Watermark image Watermarked image Extracted image
“tajmahal.jpg “lena.bmp”. PSNR= Inf Ncc= 1,Ber= 0

Original image “tajmahal.jpg Gray scale Watermark image Watermarked image Extracted image
“Lena.jpg”. PSNR=45.2629 Ncc= 0.9992,Ber= 0.1345

salt pepper noise, Rotation, Gaussian Noise,
In table 4, the proposed method is compared
Sharpening, Row and Colum removal and Row
with contourlet and SVD based algorithm [11].It
and column copying.
demonstrates that proposed method is superior to

10
109 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
TABLE 4: COMPARISON OF CT+SVD AND 7. REFERENCES:
NSCT + SVD
[1]. C.Venkata Narasimhulu &K.Satya Prasad:”A novel robust
watermarking technique based on nonsubsampled contourlet
S.No ATTACK Normalized Correlation transform and SVD”, International Journal of multimedia and
NSCT+SVD CT+SVD Applications.vol.3, no.1, Feb2011.
1 Jpeg 0.9985 0.9996
compression [2]. C.Venkata Narasimhulu &K.Satya Prasad:”A hybrid
2 Jpeg2000 0.9995 0.9996 watermarking scheme using contourlet transform and
3 Salt & pepper 0.6948 0.6823 singular value Decomposition”, IJCSNS: International
Journal of Computer Science and Network Security.vol.10,
noise no.9, Sep2010.
4 Low pass 0.9729 0.9839
filtering [3] Minh N. Do, and Martin Vetterli, “The Contourlet
5 Automatic 0.9538 0.9658 Transform: An Efficient Directional Multiresolution Image
cropping Representation” IEEE Transaction on image processing, vol
6 Histogram 0.9808 0.9733 14, issue no 12, pp 2091-2106, Dec 2005
Equation
[4] Jianping Zhou; Cunha, A.L, M.N.Do, “Nonsubsampled
7 Rotation 0.9958 0.9750 contourlet transform construction and application in
8 Median 0.9484 0.9680 enhancement” IEEE Trans. Image Proc Sept. 2005.
filtering
9 Contrast 0.9985 0.9991 [5] Arthur L. Cunha, J. Zhou, and M. N. Do, “Nonsubsampled
adjustment contourlet transform: filter design and applications in
10 Weiner filter 0.9982 0.9989 denoising” IEEE International conference on image
processing, September 2005.
11 Gamma 0.9989 0.9995
correction [6] V.I.Gorodetski L.J.Popyack, and V.Samoilov, “SVD-based
12 Gaussian Noise 0.8399 0.7538 approach to transparent embedding data into digital
13 Sharpening 0.8379 0.8212 images,” in proc. int. Workshop, MMM-ACNS,
14 Gaussian 0.9719 0.9841 St Peterburg, Russia, May 2001, pp.263-274.10.
Blurring
[7] R.Liu and T.Tan, “An SVD-Based Watermarking scheme
14 Shearing 0.9744 0.9857 for Protecting rightful ownership,” IEEE Trans. Multimedia,
16 dilatations 0.9443 0.9678 vol.4, no.1, pp.121-128, Mar.2002.
17 Color to 0.8163 0.8693
grayscale [8] A. L. Cunha, J. Zhou, and M. N. Do, “The Nonsubsampled
18 Row & Colum 0.9977 0.9972 contourlet transform: theory, design and applications,”
removal IEEE Trans. Image Proc., vol.15, no.10, October 2006.
19 Row Colum 0.9902 0.9820
[9] Emir Ganic and ahmet M. Eskicioglu “ Robust embedding
copying of visual watermarks using discrete wavelet transform and
20 Scaling (150%) 0.9187 0.9417 singular value decomposition Journal. Of Electron.
Imaging, Vol. 14, 043004 (2005); doi:10.1117/1.2137650
Published 12 December 2005
6. CONCLUSION:
[10] Dongyan liu,wenbo Liu,Gong Zhang,”An adaptive
In this paper, a novel robust hybrid watermarking watermarking scheme based on nonsubsampled contourlet
scheme is proposed for authentication of color transform for color image authentication”.Proceedings of
images using nonsubsampled contourlet transform the 2008 the 9th international conference for Young
computer Scientist,ISBN:978-0-7695-3398-8.
and singular value decomposition. Watermark is
embedded in all color spaces of host image by [11] C.Venkata Narasimhulu &K.Satya Prasad:”A new SVD
modifying singular values of mid frequency sub band based hybrid color image watermarking for copy right
coefficients with respect to watermark mid frequency \ protection using Contourlet transform”, Communicated
to International Journal of computer and
sub band coefficient with suitable scaling factor. The Applications(IJCA) in March 2011.
robustness of watermark is improved for common
image procession operations by combining both the [12] Ashraf. K. Helmy and GH.S.El-Taweel “Authentication
concepts of nonsubsampled contourlet transform and Scheme Based on Principal Component Analysis for
Satellite Images” International Journal of Signal
singular value decomposition. The proposed Processing, Image Processing and Pattern Recognition
algorithm is tested for different host images and Vol. 2, No.3, September 2009.
respective watermark images are obtained without
any visual degradation. The proposed algorithm [13] Matlab 7.6 version, Image Processing Tool Box.
preserves high perceptual quality of the watermarked
image and shows an excellent robustness to attacks
like Salt and Pepper Noise, Gaussian Noise, Row
Column Copying, and Row Column Removal.

11
110 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011
AUTHORS PROFILE:

K.Satya Prasad
C.V Narasimhulu
Received his Ph.D degree from IIT Madras, India. He
He received his Bachelor degree in Electronics and is presently working as professor in ECE department,
Communication Engineering from S.V. University, JNTU college of Engineering Kakinada and Rector of
Tirupati, India in 1995 and Master of Technology in JNT University, Kakinada, India. He has more than
Instruments and Control Systems from Regional 30 years of teaching and research experience. He
Engineering College Calicut, India in 2000.He is published 30 research papers in international and 20
currently pursuing the Ph.D degree in the department research papers in National journals. He guided 8
of Electronics and Communication Engineering from Ph.D thesises and 20 Ph.D thesises are under his
Jawaharlal Nehru Technological University guidance. His area of interests is digital signal and
Kakinada, India. He has more than 15 years image processing, communications, adhoc networks
experience of teaching under graduate and post etc..,
graduate level. He is interested in the areas of signal
processing and multimedia security

12
111 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
Parallel Implementation of Compressed Sensing
Algorithm on CUDA- GPU
Kuldeep Yadav1, Ankush Mittal 2 M. A. Ansar3, Avi Srivastava4
1, 2, 4
Computer Science and Engineering Galgotia College of Engineering
College of Engineering Roorkee Gr. Noida, INDIA
Roorke-247667, INDIA ma.ansari@ieee.org3
Kul82_deep@rediffmail.com1 aviarodonix12@yahoo.com4
dr.ankush.mittal@gmail.com2

Abstract - In the field of biomedical imaging, compressed The basis principle is that sparse or compressible signals can
sensing (CS) plays an important role because compressed be reconstructed from a surprisingly small number of linear
sensing, a joint compression and sensing process, is an measurements, provided that the measurements satisfy an
emerging field of activity in which the signal is sampled and
incoherence property. Such measurements can then be
simultaneously compressed at a greatly reduced rate.
regarded as a compression of the original signal, which can
Compressed sensing is a new paradigm for signal, image and
be recovered if it is sufficiently compressible. A few of the
function acquisition. In this paper we have worked on Basis
Pursuit Algorithm for compressed sensing. We have computed many potential applications are medical image
time for running this algorithm on CPU with Intel® reconstruction [5], image acquisition [6], and sensor
Core™2Duo T8100 @ 2.1GHz and 3.0 GB of main memory networks [7]. The first algorithm presented in this context is
which run on Windows XP. The next step was to convert this known as basis pursuit [8]. Let Φ be an M × N measurement
code in GPU format i.e. to run this program on GPU NVIDIA matrix, and Φx = b the vector of M measurements of an N
GeForce series 8400m GS model having 256 MB of video
dimensional signal x. The reconstructed signal u∗ is the
memory of DDR2 type and bus width of 64bit. The graphic
driver we used is of 197.15 series of NVIDIA. Both the CPU minimizer of the L1 Norm subject to the data min ||u||= 1,
and GPU version of algorithm is being implemented on the subject to Φu = b .A remarkable result of Candes and Tao
Matlab R2008b. The CPU version of the algorithm is being [9] is that if, for example, the rows of Φ are randomly
analyzed in simple Matlab but the GPU version is being chosen, Gaussian distributed vectors, there is a constant C
implemented with the help of intermediate software JACKET such that if the support of x has size K and M ≥ CK log
V1.3. For using Jacket, we have to make some changes in our
(N/K), then the solution will be exactly u∗ = x with
source code so to make the CPU and GPU to work
simultaneously and thus reducing the overall computational overwhelming probability. The required C depends on the
acceleration of the algorithm. Graphic Processing Units desired probability of success, which in any case tends to
(GPUs) are emerging as powerful parallel systems at a cheap one as N →∞. Donoho and Tanner [10] have computed
cost of a few thousand rupees. We got the speed up around 8X, sharp reconstruction thresholds for Gaussian measurements,
for most of the Biomedical images and six of them have been so that for any choice of sparsity K and signal size N, the
included in this paper, which can be analyzed via the profiler.
required number of measurements M to recover x can be
determined precisely.In this study, we implemented Basis
Keywords — Compressive sensing, Basis Pursuit Algorithms,
Pursuit Algorithms, on NVIDIA’s GeForce 8400 GPU with
Jacket v 1.3, GPU, medical image processing; high performance
computing.
the Computer Unified Device Architecture(CUDA)

I. INTRODUCTION programming environment. Hence, we have chosen to
implement it and we hope that other GPGPU researchers in
Current papers [1, 2, 3, and 4] have introduced the concept
the field will also make the same choice to standardize the
known as compressed sensing (among other related terms).
performance comparisons.

112 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
II. BACKGROUND A. Jacket Overview
In this section, we have discussed about the Jacket v 1.3 Jacket connects Matlab to the GPU. Matlab is a technical
which is a Graphics Processors for general purpose computing language that integrates computation,
computing. Over the past few years, specialized visualization and programming in an easy to use
coprocessors from floating point hardware to field environment that has found wide popularity both in the
programmable gate arrays have enjoyed a widening industry and academia. It is used across the breath of
performance gap with traditional x86 based processors. Of technical computing applications including mathematical
these, graphics processing units (GPUs) have advanced at an computations, algorithm development, data analysis, data
astonishing rate, currently capable of delivering over 1 visualization and application development. With the GPU as
TFOPS of single precision performance and over 300 a backend computation engine, Jacket brings together the
GFLOPS of double precision while executing up to 240 best of three important computational worlds: computational
simultaneous threads in one low cost package. As such, speed, visualization, and the user friendliness of M
GPUs have gained significant popularity as powerful tools programming. Jacket enables developers to write and run
for high performance computing (HPC) achieving 20100 code on the GPU in the native M language used in Matlab.
times the speed of their x86 counterparts in applications Jacket accomplishes this by automatically wrapping the M
such as physics simulation, computer vision, options language into a GPU compatible form. By simply casting
pricing, sorting, and search. As with previous research input data to Jacket’s GPU data structure, Matlab functions
compressed sensing studies based on Graphics Processing are transformed into GPU functions. Jacket also preserves
Units (GPUs) provide fast implementations. However, only the interpretive nature of the M language by providing
a small number of these GPU-based studies concentrate on realtime, transparent access to the GPU compiler.
compressed sensing Since the GPU which we have taken
B. Integration with Matlab
(NVIDIA 8400M GS) is the most basis model has high
Once Jacket is installed, it is transparently integrated with
portability and is easily available in present day laptop and
the Matlab’s user interface and the user can start working
desktops so can be implemented directly. However
interactively through the Matlab desktop and command
synchronizing of host and device with suitable parallel
window as well as write M-functions using the Matlab
implementation is the most challenging part. Which has
editor and debugger. All Jacket data is visible in the Matlab
been parallelized by us? Basis Pursuit Algorithms cannot be
workspace, along with any other Matlab matrices.
parallelized straight forward because of distribution part, so
our solution provides balanced and data distributed
C. GPU Data Types
parallelization framework of Basis Pursuit Algorithms on
CUDA without compromising numerical precision. We Jacket provides GPU counterparts to MATLAB’s CPU data
have broken the process in threads of blocks and managed types, such as real and complex double, single, uint32,
those threads inside a special thread managing hardware int32, logical, etc. Any variable residing in the host (CPU)
called as GPU with the help of the environment and set of memory can be cast to Jacket’s GPU data types. Jacket’s
libraries provided by CUDA. memory management system allocates and manages
memory for these variables on the GPU automatically,
behind the scenes. Any functions called on GPU data will
execute on the GPU automatically without any extra

113 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
programming, GPUfunction Jacket provides the largest
available set of GPU functions in the world, ranging from A. Basis Pursuit Compression Algorithm
functions like sum, sine, cosine, and complex arithmetic to This algorithm takes an input as an image, reads it as a
more sophisticated functions like matrix inverse, singular matrix. It then decomposes the image into blocks, then from
value decomposition, Bessel functions, and Fast Fourier blocks to many columns. Each of the columns is then
Transforms. The supported set of functions continues to processed and compressed. Each of the compressed columns
grow with every release of Jacket (see the Function is reconstructed into compressed blocks. Each compressed
Reference Guide), runtime of jacket is the most advanced block is then reconstructed and sampled to get the final
GPU runtime in the world, providing automated memory compressed image. In this algorithm there are six functions
management, compile on the fly, and execution is used they are as bp_basis, bp_decompose, bp_construct,
optimizations for Jacket enable code, Jacket’s Graphics bp_block_decompose, bp_block_construct, imagproc. These
Toolbox is the only tool in the world that enables a merger functions are collectively used to decompose the image into
of GPU visualizations with computation. With Jacket a blocks, compress the image, and reconstruct it by sampling
simple graphics command can be added at the end of a it again. The general scheme of algorithm is shown below.
simulation loop to visualize data as it is being computed
while maintaining performance, The Developer SDK makes /* Pseudo Code of Basis Pursuit Algorithm */
integration of custom CUDA code into Jacket’s runtime
1. Load real time RGB image.
very easy. With a few simple SDK functions, your CUDA
2. Assign it with double precision matrix.
code can benefit from the optimized Jacket platform. When
3. repeat
Jacket applications have completed the development, test,
for each row of block
and optimization stages and are ready for deployment, the
4. Using CUDA threads Decompose coloumnwise by basis
Jacket MATLAB Compiler allows users to generate license
pursuit algorithm.
free executables for distribution to larger user bases. (See
Find the decomposed matrix.
the SDK and JMC Wiki pages) and Interactive help for any
End for.
Jacket function is available using Jacket’s ghelp function.
5. Until all three colour block decompose.
6. repeat

III. IMPLEMENTATION for each row of block
7. Parallely using CUDA threads Reconstruct coloumnwise
In this section, first, we introduce the general scheme for by basis pursuit algorithm.
Basis Pursuit algorithm. Then, we introduce our GPU Consider complex matrix also.
implementation environment by first discussing why GPUs Find the reconstructed matrix.
are a good fit for medical imaging applications and then End for.
presenting NVIDIA’s CUDA platform and GeForce 8400 M 8. until all three color block reconstructed.
architecture. Next, we talk about the CPU implementation 9. Combine all three color block to give a matrix.
environment. This is followed by description of the test data 10. convert double precision of compressed
used in the experiments. Finally, we provide the list of Image data to unsigned int values.
CUDA kernels used in our GPU implementation. 11. Scale image data

114 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
B. GPU Implementation Environment fine, lightweight threads in parallel. In CUDA, programs are
We have implemented a GPU version of Basis pursuit (BP) expressed as kernels. Kernels have a Single Program
with NVIDIA’s GPU programming environment, CUDAv Multiple Data (SPMD) programming model, which is
0.9. The era of single-threaded processor performance essentially Single Instruction Multiple Data (SIMD)
increases has come to an end. Programs will only increase in programming model that allows limited divergence in
performance if they utilize parallelism. However, there are execution. A part of the application that is executed many
different kinds of parallelism. For instance, multicore CPUs times, but independently on different elements of a dataset,
provide task-level parallelism. On the other hand, GPUs can be isolated into a kernel that is executed on the GPU in
provide data-level parallelism. Depending on the application the form of many different threads. Kernels run on a grid,
area, the type of the preferred parallelism might change. which is an array of blocks; and each block is an array of
Hence, GPUs is good fit for all problems. However, medical threads. Blocks are mapped to multiprocessors within the
imaging applications are very suitable to be implemented on G80 architecture, and each thread is mapped to single
GPU architecture. It is because these applications processor. Threads within a block can share memory on a
intrinsically have data-level parallelism with high compute multiprocessor. But two threads from two different blocks
requirements, and GPUs provide the best cost-per- cannot cooperate. The GPU hardware performs switching of
performance parallel architecture for implementing such threads on multiprocessors to keep processors busy and hide
algorithms. In addition, most medical imaging applications memory latency. Thus, thousands of threads can be in flight
(e.g. semi-automatic segmentation) require, or benefit from at the same time, and CUDA kernels are executed on all
visual interaction and GPUs naturally provide this elements of the dataset in parallel. We would like to
functionality. Hence, the use of the GPU in non-graphics mention that in our implementation; increasing the dataset
related highly-parallel applications, such as medical imaging size does not have an effect on the shared memory usage.
applications, became much easier than before. For to the This is because, to deal with larger datasets, we only
graphics API. Since it is cumbersome to use graphics APIs increase the number of blocks and keep the shared memory
for non-graphics tasks such as medical applications, allocations in a thread as well as the number of threads in a
instance, NVIDIA introduced CUDA to perform data- block the same.
parallel computations on the GPU without the need of C. CPU Implementation Environment
mapping the graphics-centric nature of previous
The CPU version of basis pursuit is implemented in Matlab
environments. GPU- specific details and allowing the
with integration of jacket v1.3. The computer used for the
programmer to think in terms of memory and math
CPU implementation is an Intel® CORE™2Duo T8100 @
operations as in CPU programs, instead of primitives,
2.1GHz and 3.0 GB of main memory which run on
fragments, and textures that are specific to graphics
Windows XP (SP2). The CPU implementation was executed
programs. CUDA is available for the NVIDIA GeForce
on this box with both single threading mode and multi-
8400 (G80) Series and beyond. The GeForce 8400 GS
threading mode. Open MP is used to implement the multi-
model having 256Mbytes of video memory of DDR2 type
threading part.
and bus width of 64bit l. The memory bandwidth of the
GeForce 8400 GTX is 80+ GB/s. To get the best
performance from G80 architecture, we have to keep 128
processors occupied and hide memory latency. In order to
D. Test Data of (BP) Compressed Results
achieve this goal, CUDA runs hundreds or thousands of

115 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
Our test data consists of six MRI images which measure the implementations. It is clear from the figures that significant
performance of images on CPU & GPU with the help of speedup is obtained for Basis Pursuit in GPU in comparison
profiler and has been shown in Figure 2, 3, 4 below. Show to CPU implementations.
the profiler results for CPU and GPU based

(a) MRI1 (b) MRI 2 (c) MRI 3

(d) MRI 4 (E) MRI 5 (F) MRI 6

Figure 1: Test Data Images Under Measure the Performance

116 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
Figure 2: Profile on CPU

Figure 3: Profile on GPU

5. Experiment and Results
Sharpetal’s implementation and highlight our
In this section, we first compare the runtimes of our GPU
improvements. We measure the performance of images on
and CPU implementations for datasets with different sizes.
CPU & GPU with the help of profiler and have been shown
And present our speedup. Then, we show visual results by
in figure 2 and 3 show the profiler results for CPU
providing slices from one of the datasets. Next, we provide
architecture and GPU based implementations. It is clear
the breakdown of GPU implementation runtime to the
from the figures that significant speedup is obtained. We
CUDA kernels and present GFLOP (Giga Floating Point
have achieved around 8X speedup over a single-threaded
Operations = 109 FLOP) and GiB (Gibi Bytes = 230 Bytes)
CPU implementation over GPU. We can calculate the speed
instruction and data throughput information. This is
up by using the following formula. Speed up=time taken to
followed by the discussion of the factors that limit our
compress the image on CPU/time taken to compress image
performance. Finally, we compare our implementation with
on GPU.

117 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
Table1. Performance of GPU implementation with respect to CPU implementation

Time to run Time to run Time to run
S. No. Image
Size program on BP_DECOMPOSE BP_DECOMPOSE Depth
CPU on CPU(sec) on GPU(sec)
1. MRI1 256 X 256 118.52 105.2 14.06 512
2. MRI2 256 X 256 136.47 123.49 20.31 1024
3. MRI3 256 X 256 139.56 125.159 18.19 1024
4. MRI4 256 X 256 98.87 84.25 11.90 512
5 MRI5 256 X 256 120.34 106.1 16.31 512
6 MRI6 256 X 256 138.47 124.50 21.02 1024

Figure 4: Comprarative Graphical Results on CPU& GPU

6. CONCLUSION

The work presented in this paper is one of the major
problems of clinical application of compressed sensing on
bio-medical imaging which is its high computational cost by
using currently available jacket software and GPU NVIDIA
GeForce series 8400m GS model having 256M bytes of
Video memory of DDR2 type and bus width of 64bit. We
reduced the execution time from 50 seconds to 6 seconds for
several bio-medical images. We obtained the speedup of
the system 8x time.

118 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
REFERENCES

[1] Rick chartrand and wotayin “Iteratively Reweighted algorithms for
Compressed Sensing”, in IEEE 2008 ICASSP 2008.
[2] D. L. Donoho, “Compressed sensing,” IEEE Trans.On Information
Theory, vol. 52, no. 4, pp. 1289–1306, April 2006.
[3] http //www.dsp.ece.rice.edu / cs.
[4] E. J. Cand`es, J. Romberg, and T. Tao, “Robust uncertainty principles:
[[[[[

Exact signal reconstruction from highly incomplete frequency
information,” IEEE Trans. Inf. Theory, vol. 52, 2006.
[5]. M. Lustig, J. M. Santos, J.-H. Lee, D. L. Donoho, and J. M. Pauly,
“Application of compressed sensing for rapid MR imaging,” in
SPARS, (Rennes, France), 2005.
[6] D. Takhar, J. N. Laska, M. B. Wakin, M. F. Duarte, D. Baron, S.
Sarvotham, K. F. Kelly, and R.G. Baraniuk, “A new compressive
imaging camera architecture using optical-domain compression,” in
Computational Imaging IV - Proceedings of SPIE-IS and T
Electronic Imaging, vol. 6065, 2006.
[7] M. F. Duarte, S. Sarvotham, D. Baron, M. B. Wakin, and R. G.
Baraniuk, “Distributed compressed sensing of jointly sparse
signals,” in 39th Asilomar Conference on Signals, Systems and
Computers, pp. 1537–1541, 2005.
[8] S. S. Chen, D. L. Donoho, and M. A. Saunders, “Atomic
decomposition by basis pursuit,” SIAM J. Sci. Comput., vol. 20, pp.
33–61, 1998.
[9] E. Cand`es and T. Tao, “Near optimal signal recovery from random
projections: universal encoding strategies,” IEEE Trans. Inf.
Theory, vol. 52, pp. 5406–5425, 2006.
[10] D. L. Donoho and J. Tanner, “Thresholds for the recovery of sparse
solutions via L1 minimization,” in 40th Annual Conference on
Information Sciences and Systems, pp. 202–206, 2006.

119 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011

FUZZY HRRN CPU SCHEDULING
ALGORITHM
1
Bashir Alam, 1 M.N. Doja, 2R. Biswas, 3M. Alam
1
Department of Computer Engineering, Jamia Millia Islamia, New Delhi, India
2
Department of Computer Science and Engineering, Manav Rachna University, Faridabad, India
3
Department of computer Science, Jamia millia Islamia,New Delhi India
Email: - babashiralam@gmail.com, ndoja@yahoo.com , ranjitbiswas@yahoo.com, mansaf@lycos.com

Abstract— There are several scheduling algorithms like FCFS, (iii) SRTN: In shortest Remaining time next scheduling
SRTN, RR, priority etc. Scheduling decisions of these algorithms algorithm , the process with shortest remaining
are based on parameters which are assumed to be crisp.
However, in many circumstances these parameters are vague. time is scheduled for execution.[3]
The vagueness of these parameters suggests that scheduler should (iv) Priority: in Priority Scheduling algorithm the process
use fuzzy technique in scheduling the jobs. In this paper we have with highest priority is scheduled for execution.
proposed a novel CPU scheduling algorithm Fuzzy HRRN that
(v) Round-robin: In this the CPU scheduler goes around
incorporates fuzziness in basic HRRN using fuzzy Technique FIS.
the ready queue allocating the CPU to each
Keywords: - HRRN, CPU Scheduling, FIS, Fuzzy Logic process for a time interval of up to one time
quantum. [1,2,3]
1. INTRODUCTION
(vi) Multilevel queue scheduling: In this the ready queue
When a computer is multi programmed, it frequently has is partitioned into several separate queue. The
multiple processes competing for the CPU at the same time. processes are permanently assigned to one queue
When more than one process is in the ready state and there is generally based on some property of the process
only one CPU available, the operating system must decide such as memory size, process priority or process
which process to run first. The part of operating system that type. Each queue has its own scheduling
makes the choice is called short term scheduler or CPU algorithm. There is scheduling among the
scheduler. The algorithm that it uses is called scheduling
queues, which is commonly implemented as
algorithm. There are several scheduling algorithms. Different
scheduling algorithms have different properties and the choice fixed-priority preemptive scheduling. Each
of a particular algorithm may favor one class of processes over queue has absolute priority over low priority
another. Many criteria have been suggested for comparing queues.[1]
CPU scheduling algorithms and deciding which one is the best (vii) Multilevel feedback-queue scheduling:-This allows a
algorithm [1]. Some of the criteria include the following:- process to move between queues.[1]
(i)Fairness (viii) Fair share Scheduling: Fair share scheduler
(ii)CPU utilization
considers the execution history of a related group
(iii)Throughput
(iv)Turnaround time of processes, along with the individual execution
(v)Waiting time history of each process in making scheduling
(vi)Response time decision. The user community is divided into a
It is desirable to maximize CPU utilization and throughput, to fair- share groups. Each group is allocated a
minimize turnaround time, waiting time and response time and fraction of CPU time. Scheduling is done on the
to avoid starvation of any process. [1, 2]
basis of priority of the process, its recent
Some of the scheduling algorithms are briefly described
below: processor usage and the recent processor usages
(i) FCFS: In First come First serve scheduling algorithm of the group to which the process belongs. Each
the process that request first is scheduled for process is assigned a base priority. The priority
execution. [1,2,3] of a process drops as the process uses the
(ii) SJF: In shortest Job first scheduling algorithm the processor and as the group to which process
process with the minimum burst time is belongs uses the processor.[3]
scheduled for execution.[1,2]

120 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
(ix) Guaranteed scheduling:-In this a ratio of actual CPU "consequent". Typical fuzzy inference subsystems have
time a process had and its entitled CPU time is dozens of rules. These rules are stored in a knowledgebase. An
calculated. The process with this lowest ratio is example of fuzzy IF-THEN rules is: IF Remaining Time is
extremely short then priority is very high, in which Remaining
scheduled[2]
Time and priority are linguistics variables and extremely short
(x) Lottery Scheduling:-The basic idea is to give and very high are linguistics terms. The five steps toward a
processes lottery tickets for CPU time. Whenever fuzzy inference are as follows:
a scheduling decision has to be made , a lottery • fuzzifying inputs
ticket is chosen at random and the process • applying fuzzy operators
holding the ticket gets the CPU.[2] • applying implication methods
(xi) HRRN: - In this response ration is calculated for each • aggregating outputs
• defuzzifying results
process. The process with the highest ratio is
Below is a quick review of these steps. However, a detailed
scheduled for execution. [3] study is not in the scope of this paper. Fuzzifying the inputs is
the act of determining the degree to which they belong to each
In all the above scheduling algorithm the parameters used are of the appropriate fuzzy sets via membership functions. Once
crisp. However, in many circumstances these parameters are the inputs have been fuzzified, the degree to which each part
vague. [9] To exploit this vagueness we have used fuzzy logic of the antecedent has been satisfied for each rule is known. If
in our proposed scheduling algorithm. We have also done the antecedent of a given rule has more than one part, the
simulation for comparing the performance of basic HRRN fuzzy operator is applied to obtain one value that represents
scheduling algorithm and Fuzzy HRRN scheduling algorithm. the result of the antecedent for that rule. The implication
function then modifies that output fuzzy set to the degree
2. FUZZY INFERENCE SYSTEMS AND FUZZY specified by the antecedent. Since decisions are based on the
LOGIC testing of all of the rules in the Fuzzy Inference System (FIS),
A fuzzy inference system (FIS) tries to derive answers from a the results from each rule must be combined in order to make
knowledgebase by using a fuzzy inference engine. The the final decision. Aggregation is the process by which the
inference engine which is considered to be the brain of the fuzzy sets that represent the outputs of each rule are processes
expert systems provides the methodologies for reasoning into a single fuzzy set. The input for the defuzzification
around the information in the knowledgebase and formulating process is the aggregated output fuzzy set and the output is
the results. Fuzzy logic is an extension of Boolean logic then a single crisp value [4]. This can be summarized as
dealing with the concept of partial truth that denotes the extent follows: mapping input characteristics to input membership
to which a proposition is true. Whereas classical logic holds functions, input membership function to rules, rules to a set of
that everything can be expressed in binary terms (0 or 1, black output characteristics, output characteristics to output
or white, yes or no), fuzzy logic replaces Boolean truth values membership functions, and the output membership function to
with the degree of truth. Degree of truth is often employed to a single crisp valued output. There are two common inference
capture the imprecise modes of reasoning that play an methods [4]. The first one is called Mamdani's fuzzy inference
essential role in the human ability to make decisions in an method proposed in 1975 by Ebrahim Mamdani [5] and the
environment of uncertainty and imprecision. The membership second one is Takagi-Sugeno-Kang, or simply Sugeno,
function of a fuzzy set corresponds to the indicator function of method of fuzzy inference introduced in 1985 [6]. These two
the classical sets. It can be expressed in the form of a curve methods are the same in many respects, such as the procedure
that defines how each point in the input space is mapped to a of fuzzifying the inputs and fuzzy operators. The main
membership value or a degree of truth between 0 and 1. The difference between Mamdani and Sugeno is that the Sugeno’s
most common shape of a membership function is triangular, output membership functions are either linear or constant but
although trapezoidal and bell curves are also used. The input Mamdani’s inference expects the output membership functions
space is sometimes referred to as the universe of discourse [4]. to be fuzzy sets. Sugeno’s method has three advantages.
Fuzzy Inference Systems are conceptually very simple. An Firstly, it is computationally efficient, which is an essential
FIS consists of an input stage, a processing stage, and an benefit to real-time systems. Secondly, it works well with
output stage. The input stage maps the inputs, such as optimization and adaptive techniques. These adaptive
deadline, execution time, and so on, to the appropriate techniques provide a method for the fuzzy modeling procedure
membership functions and truth values. The processing stage to extract proper knowledge about a data set, in order to
invokes each appropriate rule and generates a result for each. compute the membership function parameters that best allow
It then combines the results of the rules. Finally, the output the associated fuzzy inference system to track the given
stage converts the combined result back into a specific output input/output data. The third, advantage of Sugeno type
value [4]. As discussed earlier, the processing stage, which is inference is that it is well-suited to mathematical analysis.
called the inference engine, is based on a collection of logic The block diagram of the proposed fuzzy inference system is
rules in the form of IF-THEN statements, where the IF part is given in figure1.
called the "antecedent" and the THEN part is called the

121 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
In the proposed model, the input stage consists of three Figure 2: Memebership Function for Static Priority
linguistic variables. The first one is the static priority that is
assigned to the process before its execution. The second is the
expected remaining time of the process. The third input is the
Response Ratio of the process. The output stage consists of
one linguistic variable called Dynamic priority.
Static Priority
Fuzzy
Inference Dynamic
Remaining Engine Priority
Time (Sugeno)
27 Rules Figure 3: Membership function for Remaining Time
Response
Ratio

Figure 1: Block diagram of FIS

The input and out variables are mapped into fuzzy sets using
appropriate membership functions. The shape of the
membership function for each linguistic term is determined by
the expert. Adjusting these membership functions in an
optimal mode is very difficult. However, there are some
techniques for adjusting membership functions [7,8]These
techniques cannot be covered in this paper. They can be
further studied in a separate paper. Figure 4: Membership Function for Response Ratio
The membership functions for inputs and outputs are given
below
Membership Function for DP (Dynamic Priority)

Type- Triangular, Range:1-5, Very low-[-1,0,1], Low:-
[0,1.5,3] medium:-[2,3,4] High:-[3,4,5] Very High:-[4,5,6]

Membership Function for SP (Static Priority)
5
Type- Triangular, Range: 1-5, low-[-2, 0, 2], medium-[1, 2.5, Figure5: Membership Function for Dynamic Priority
4] High [3, 5, 7] Twenty seven rules are formulated and a Sugeno type fuzzy
Inference system is built. Some of the rules are listed below:
Membership Function for RT (Remaining Time) • If the static priority is
‘low’ and remaining time is ‘extremely short’ and
Type- Triangular, Range: 0-5, Extremely short:-[-2, 0, 2], Response Ratio is ‘long’ then the dynamic priority is
Very Short:-[1, 2.5, 4] Short:-[3, 5, 7]
‘very high’.
Membership Function for RR (Response Ratio) • If the static priority is
‘low’ and remaining time is ‘ short’ and Response
Type- Triangular, Range: 0-10, Short:-[0, 1, 2], Medium:-[2,
Ratio is ‘short’ then the dynamic priority is ‘very
5, 8] Long:-[5, 10, 15]
low’.
• If the static priority is
‘medium’ and remaining time is ‘ extremely short’
and Response Ratio is ‘long’ then the dynamic
priority is ‘very high’.
• If the static priority is
‘medium’ and remaining time is ‘short’ and Response
Ratio is ‘short’ then the dynamic priority is
‘medium’.

122 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011

Table I: Rule base for Fuzzy Inference System
S. Static Remainin Respons Dynamic
No priority g time e Ratio priority
.
1. low Extremely short Very High
short
2. Low Extremely medium Very High
short
3. low Extremely long Very High
short
4. low Very short short Very low
5. low Very short medium low
6. low Very short long high
7. low short short Very low
8. low short medium low
Figure 6: Rule View of FISHRRN
9. low short long high
10. Medium Extremely short Very high
short
11. Medium Extremely medium Very high
short
12. Medium Extremely long Very high
short
13. Medium Very short short Medium
14. Medium Very short medium Medium
15. Medium Very short long Very High
16. Medium short short Medium
17. Medium short medium medium
18. Medium short long High
19. High Extremely short Very high
short
20. High Extremely medium Very high
short
21. High Extremely long Very high
short
22. High Very short short High Figure 7: Surface view of FISHRRN
23. High Very short medium High
24. High Very short long Very High 4. Proposed Algorithm
25. High short short high
26. High short medium High The parameters of process are stored in table called Process
27. High short long Very High Control Block (PCB). Each process has its own PCB. The
structure of the Process Control Block is given in Table II.
Table II: Structure of Process control Block The parameters remaining time Rti, static priority spi , dynamic
priority dpi and waiting time wti of process Pi are stored in
Process Process Control Block PCBi . Response Ratio RRi of a
Process name=”bash” process Pi may be calculated by dividing the sum of waiting
Process identifier=100 time and service time by service time. The proposed algorithm
State=Ready is as follows:
CPU reserve i. For each process Pi in ready queue fetch its
{CPU Burst Time= parameters Rti, spi, and wti from PCBi and
CPU Remaining Time= calculate RRi and give them as input to FIS and
Priority=
then set dpi to the output of FIS
Waiting time=
Arrival Time= ii. Schedule the process Pi with the highest value of dpi
Start time= for execution.
….
}

123 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
iii. If the scheduled process finishes and no new request [8] Simon D, Training fuzzy systems with the extended
arrives go to step ii Kalman filter, Fuzzy Sets and Systems, Vol. 132, No. 2, 1,
iv. If new request arrives go to step I December 2002.
[9]Bashir Alam, M. N. Doja & R. Biswas, “A General Fuzzy
5. Performance Comparison CPU Scheduling Algorithm” ,International Journal of Fuzzy
Systems and Rough Systems (IJFSRS) (Vol. 1, No. 1, Janu.-
For comparing the performance of basic HRRN Scheduling June 2008)
and Fuzzy HRRN we did simulation on 1000 processes in
groups of ten each. We assumed random burst time of
processes and random arrivals. Max burst time of a process
should not exceed 10 ms. Throughput and average waiting
time of the processes in a group was computed and then
average was taken over all groups to give average throughput
and average waiting time. The performance is shown in the
column chart given in figure given below. This chart shows
that the average waiting time for Fuzzy HRRN is lesser than
the same of basic HRRN. This also shows that throughput of
our proposed algorithm Fuzzy HRRN is better than the same
of basic HRRN.

Figure 8 : Performance Comparison of HRRN and Fuzzy
HRRN Scheduling

6. Conclusion

Our proposed algorithm named Fuzzy HRRN has got benefits
of shortest remaining time next (SRTN) as well as HRRN
scheduling algorithm and fuzziness. This proposed algorithm
gives better throughput and lesser average waiting time than
its non fuzzy counter algorithm HRRN.

REFERENCES

[1] Silberschatz, A., Peterson, J. L., and Galvin, .B.,Operating
System Concepts, Addison Wesley, 7th Edition, 2006.
[2] Andrew S. Tanenbaum , and Albert S. Woodfhull, Opera-
ting Systems Design and Implementation,Second Edition,2005
[3]William Stallings, Operating Systems Internal and Design
Principles, 5th Edition ,2006
[4]Wang Lie-Xin, A course in fuzzy systems and control,
Prentice Hall, August 1996.
[5]Mamdani E.H., Assilian S., An experiment in linguistic
synthesis with a fuzzy logic controller,International Journal of
Man-Machine Studies, Vol.7, No. 1, 1975.
[6] Sugeno, M., Industrial applications of fuzzy control,
Elsevier Science Inc., New York, NY, 1985.
[7]Jang, J.-S. R., ANFIS: Adaptive-Network-based Fuzzy
Inference Systems, IEEE Transactions on Systems, Man, and
Cybernetics, Vol. 23(3), 685, May 1993.

124 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Experiences and comparison study of EPC & UML
for Business Process & IS modeling
Md. Rashedul Islam Md. Rofiqul Islam Md. Shariful Alam Md. Shafiul Azam
School of Business and School of Business and School of Business and Dept. of Computer Science
Informatics Informatics Informatics and Engineering
Högskolan i Borås Högskolan i Borås Högskolan i Borås Science and Technology
Borås, Sweden Borås, Sweden Borås, Sweden University, Pabna
rashed.cse@gmail.com rana_aiub01@yahoo.com shajib004@yahoo.com Pabna, Bangladesh
shahincseru@gmail.com

Abstract— Business process modeling is an approach by which we describes the business process from the business perspective
can analyze and integrate the business process. Using the and it is so much understandable for business people.
Business Process Modeling we can represent the current and
future process of a business/organization/enterprise. The business On the other hand object oriented modeling is closely
process modeling is a prerequisite and essential implementing a related to implementation. Now a day both two types are
business or making any automation system. In this paper, we coming closer together for making efficient business
present our experience in a Business Process Modeling for information process modeling which is best for process the
organization. This paper presents detailed description about business also implementing the business information system.
business process modeling, details description about the main two
modeling language EPC and UML. This paper presented the In this paper we have discussed the comparison of two
uses, advantages, disadvantages of EPC and UML modeling main business process modeling language one is EPC and
language. Here we tried to express the experience about those another is UML. The EPC is mainly the process oriented
modeling language. This paper presents a details comparison modeling and the UML is mainly the objecting oriented
between two modeling language from the business process modeling. The EPC and UML have enough tools to represent
modeling and information system implementation point of view. any business process. Also those are very useful and easy to
understand to related people. After all every modeling language
Keywords- Business Process Modeling, Petri net, Event-driven has some advantage and disadvantage or difficulties. At the
Process Chain (EPC), Unified Modeling Language (UML), time of comparing of two modeling language we have found
Process-oriented modeling, Object-oriented modeling. some difficulties related to each other. Some are good and
understandable for some level of people and other for other
level of people.
I. INTRODUCTION
Business Process Modeling and successful development II. BUSINESS PROCESS MODELING
and implementation of business interrelated. Before thinking a
Business information system or Information system supporting Business Process Modeling (BPM) also we can
business process then the first comes to you the business call Business Process Discovery (BPD). Business process
process modeling. A business process modeling demonstrates modeling is an approach by which we can analyze and
the whole scenario of business to the all related people and also integrate the business process. Using the Business Process
increases the performance of business process. Modeling we can represent the current and future process of a
business/organization/enterprise. Business Process Modeling
Every business or organization consists of high number of figures out the whole business model and provides maximum
interlinked core processes and every core process has many sub business performance. The main outcomes of the Business
process and so many interrelated internal and external objects. Process Modeling are, add value for the customer and reduce
So modeling this business is very imported for implementing a the costs for the company to increase profit.
successful business and IT implementation.
A Business Process Model is commonly a diagram
There are many modeling approach has been developed for representing a sequence of activities and information flow. It
define a model of a business respect to organizational and represents the business in sequence start to end by events,
information system aspect. Different business process flows are actions and links or connection points. Business Process Model
different way and different types of people are involves in includes both IT processes and people processes. There are two
those business process. So different business process modeling main types of Business Process Model:
is relatively good for different business process and different
level of people. The Process oriented modeling is mainly 1. Baseline Model (Present)
Identify applicable sponsor/s here. (sponsors) 2. To Be Model (Future)

125 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

A. Some common issues with business process modeling [1] An Event-driven Process Chain (EPC) is
Managing collaborative activities within business mainly flowchart type modeling language. The EPC
process models that are derived from the is very useful for business process modeling. Also it
“transformational” approach is easy to understand for business people.
Canonical models and variability management c) Unified Modeling Language (UML)
Notations are stabilizing but methods are lagging Unified Modeling Language (UML) is a general-
Process decomposition purpose and Object Oriented Modeling Language.
o Some rules available, methodology Using UML we can create visual model for making
dependent IT system using graphic notation techniques.
o Becomes more important when coupled with
Business process execution and Web d) Dynamic Essential Modeling (DEMO)
Services Dynamic Essential Modeling (DEMO) is mainly
Managing requirements from business processes, to communication-centered organizational modeling
use cases to systems approach. DEMO is helpful for details specification
o Is the use case driven approach still needed? of behavior of participating actors.
(non question)
IT enablement focus –Human Interaction In this paper we describe the details and comparison of
Event-driven Process Chain (EPC) and Unified Modeling
Management tends to be relegated to forms driven
Language (UML).
approaches

B. Advantages of Business Process Modeling III. REFLECTION ON EPC MODELING LANGUAGE
There are so many type of advantage of business process Event-driven Process Chains (EPC) is a widely used
modeling. We can describe the advantage of business process approach for Modeling Business Process. The EPC provides
modeling in different aspect as follows: comprehensive means for modeling several aspects of a
business process [7]. The modeling approach Event-driven
a) Formalize existing process and spot needed Process Chains (EPC) [2, 3] has been developed for model
improvements business process within the ARIS framework.
Using the Business Process Modeling analyst can make an In whole business process there are so many business
overall structure of whole business in a graphical view. BPM function event activity. Using the EPC model, we can model
helps to represent all processes and data source internal and whole business process with different events and business
external objects which is understandable for any business function in sequences of events triggering business functions.
people. Also BMP help to make the business model which will The business functions are themselves the results of other
be adjustable for the future need. functions separately from initial events triggering the whole
process. Representing business process decisions and
b) Facilitate automated, efficient process flow
expanding the complex control flow, EPC control structure
BPM supports process parallelism. In business has different with connector operation “and”, “or” and “xor” can be used.
parallel activity which can perform independently. Using BPM This set of elements describes the processes, since some
we can model parallel activity. And it is possible to make an authors define a process as a succession of events and functions
efficient process flow [10, 11]. The connector may be used for split or join and before
and after of those connector will be event or function. The three
c) Increase productivity and decrease head count
connectors have twelve possibilities. In several standard
In a Business a suitable model can increase productivity software package (SAP) have used the EPC for making
and appropriate resource allocation reduce the cost. Using software for documenting the business process [4].
Business Process Model we can design the suitable process
model with resource allocation which is helpful to increase the A. ARIS – ARchitecture of Integrated Information Systems
productivity and decrease no of uses people.
ARIS (Scheer 2000) [5] stands for Architecture of
C. Business Process Modeling Language Integrated Information Systems, and denotes a methodology
for modeling business processes. EPCs are not a new method in
There are mainly 4 Business Process Modeling Approaches essence, as it contains elements of the Petri nets and GERT
(Scheer 2002) [6]. The ARIS methodology or its core
a) Petri net
technique EPCs, have been often confused with the software
Petri net is a mathematical modeling language for the tools such as ARIS Toolset (IDS Scheer 2003) [7]. Just to
description of distributed systems. Mainly Petri net is make one example, ARIS is labeled as a „modeling tool‟ in
one kind of directed bipartite graph. In the Petri Net Vernadat (2002) [8]. This confusion is probably augmented by
graph all transactions and places represented by node. the success of IDS Scheer, and it is considered to be a leader in
the BPM sector (Gartner Group 2002) [9].
b) Event-driven Process Chain (EPC)

126 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

The Event-driven Process Chains (EPC) diagram is the core diagram “Order confirmation”, “Order Tracking”, and
technique for modeling in ARIS. The ARIS divided a process “Production Planning” are the components of Function View.
in different aspects or views. The several views are (a) the
functional view, (b) the data view, or (c) the resource view. The d) Organization view:
EPC technique for modeling in ARIS makes link the all views The Organizational view contains the relationship between
in control view. We will describe the different views in users and organizational units. The Organizational units formed
following paragraph. by some user for performing some specific task. Human beings
are able to perform complex social action such as enterprise.
a) Descriptive Views But the complex action can be broken in manageable units. In
In previous paragraph we have mentioned the different the above diagram the content of Organizational views are
views. and above, the concept of ARIS lies on reducing “department” and “user”.
complexity into different views. In the following example,
Figure-1 Business Process Model with different views [11] e) Resource view:
describes the excerpt model of a business process with the The Resource view constitutes by resource of organization
concept of different views and link between views. The like Information technology components like Computer, Data
following example is a computer supported business process Server.
for processing customer orders. This business process includes
different processes, activity/function, events, users, f) The Control View: EPC
Organizational units and IT resources, Also relationship The Control view mainly the combine view of all views.
between component for describe the whole business flow. Initially all views are developed separately for reducing the
complexity. But in Control view links functions, organizational
units and data in together. Here functions, events, information
resources, and organization units are connected together into a
common context according to process flow. And this combine
Control view model is the complete EPC. Extended form of
EPC we can link additional elements such as Organizational
unit, data, product or service to the functions in an EPC
Diagram.

B. Purpose of EPC
Before thinking about Purpose of EPC, if we think about
the purpose of Business Modeling we can see that identifying
the all core and sub processes and make a suitable model and
Figure 1: Business Process Model with different views
way of business process is the main purposes of Business
Modeling. So Making a Business Model is the main purpose of
In the above diagram the whole model and components has EPC Modeling Language. Also there are some other main
been divided into individual views for reducing the complexity purposes like:
of the business model. The all components are divided in Develop business process model which is useful for
individual views on the basis of relationship of component and representing an outline of whole business.
action. The criteria of separation is that, the relationships Make a graphical method modeling which will be
between components in the same view are relatively strong and
easy understandable to the users.
relationships between components of different views are
relatively weak. The four views of this modeling result Gathering requirements in the beginning phase.
according to ARIS are describes bellow: Capture the flow of events in the business domain.
Describe in detail organizational aspects of the
b) Data view: business information systems. Make starting point for
The data view generally contains the events and statuses. identifying the actors of the system.
Events mainly the informational object which is represents the Distinguish between function and process
data. Here the events are “customer order received”, “order Identify software elements by analyzing business
confirmation prepared”. On the other hand Statuses are areas which related to business data and functions.
“customer status” and “article status” also represented by data. To make the whole organization more productivity
In data view the description of detail requirements plays an and maximize resource utilization
important role for developing Information System.
c) Function view: C. Advantage of EPC
The function view mentions the performing activity of a As mentioned before, the EPC is widely used modeling
business process. Also the overall relationship and relationship language which has so many features by which we can
between function. Sometime the function is complex. For represent any business model easily. Also it has many
reducing the complexity it can be broken down. In above advantages. After a thorough evaluation of a number of
methodologies, techniques, and tools, EPC has been selected

127 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

for modeling the Hospital case. The main reasons are the included with the “OR” connector such as wait-for-all, first-
following: come or every-time. . In wait-for-all approach the “OR”
connector will wait-for- all path which split by start “OR”
One of the main advantages of the EPC is that it is connector. In first-come approach the “OR” operator will
both powerful and easily understandable for end- trigger when the first path will complete.
users.
It is very much communicative to the different F. Business Perspectives and Views
function and process
For the beginning idea and study of process model, EPC
EPC is not too much technical take a part in an important role for business from business
The EPCs match the requirements posed with respect perspective and views. For designing a complex system
to the ease of understanding by non-specialist in perspectives have proven it for differentiate it. In business
modeling. process modeling, there are several perspective proposals in
EPCs can offer a multi-level view of the process, EPC. The perspectives are data oriented, application oriented,
since a function in an EPC can be described in more function oriented, organization oriented and product/ service
detail by means of another EPC. oriented.
EPCs offer a consistent, formally supported model
(see comments above) that can ensure an efficient IV. REFLECTION OF UML MODELING LANGUAGE
simulation of the processes. Modeling is an essential part of analyzing or engineering a
A workflow oriented modeling connected to actors, business, also it is very important and prerequisite for
documents and information. developing an IT system. A complex business process is
For getting clear idea about actors and function in use difficult for you to describe in textually format without a
cases EPCs can be used for explain accurately to the modeling diagrams. There are three key benefits of Modeling
workflows of use cases. are visualization, complexity management and clear
In EPCs, there is an option for translating the activity communication.
diagram into EPC and vice versa. In several modeling language the UML is most uses
EPCs are often used for capturing and discussing modeling language in all over the world. The UML is a general
business processes with people who have never been purpose visual language which is very useful for specifying,
trained in any kind of modeling technique, e. g. with constructing, documenting the complex systems [13]. Using
workers on the shop floor. UML you can make a specific model which will be
Although EPCs can be understood even by short-time unambiguous and complete way.
trained personnel, the same models can be refined
The standard UML approved by the OMG™ in 1997. In
and used for the requirements definition of an past few years there have been some improvement and UML 2
information system. is the major revision [12]. UML mainly is an Object Oriented
Modeling Language. From the developer point of view,
D. Difficulties of EPC Modeling of a business using the UML visualization graph can
We have not found much difficulty for using of EPC. There express a common interpretation at the time of exchanging
is some minor limitation in EPC language. They are: ideas. UML is not a programming Language but it helps to
interpret the business process in a way which is very helpful for
A simple EPC diagram can analyze easily, but for making IT system. Also UML modeling method cover the end-
complex graphs it is very important to analyze before users‟ views [14]. The UML also documents the project during
implementation and it is difficult. Sometime in a software development.
complex diagram a deadlock can be happened when
executing the process according to the diagram even A. System Architectural Views and UML diagrams
the graph is semantically correct according to the A Business process or IT system can be viewed from a
definition of EPC number of perspectives. There are so many categories of
Looping is another possible discovered problem. participants involved in a business process of IT system like
For simulating big model the OR operator has been Business people, User, Analysts, Project Managers, Developers
forbidden. etc in different ways and different times over the project‟s life.
So from the architecture point of view a system has different
E. Ambiguity of the OR connector viewpoints focused on a particular aspect of that system. Such
In EPC there are three logical connector “AND”, “OR” and as:
“XOR”. The using of “AND” and “XOR” is no problem. But 1. Use case view
the use of “OR” connector make some ambiguities. Which path 2. Design view
we should follows. At the end part we will wait for one path or 3. Process view
both parts and so on. The ambiguity of or connector discussed 4. Implementation view
in different approaches. “OR” connector‟s ambiguity has been
5. Deployment view
described in different way [20]. A comment flags can be

128 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

The figure 2 illustrates, the architecture of a software system can be described by five views.

Figure 2: Modeling System Architecture (From [15])

The UML has different diagram cover the static and State Diagram.
dynamic aspect/views of system. Such as the Use Case diagram
represent the use case view. The class diagram of UML Just like interaction diagram, state diagrams also
represents the design and process view. The component capture the behavior of a system over time. The
diagram represents the implementation view. The deployment elements of State Diagram are State, Transition
diagram is responsible for deployment view. In all these views Activity Diagram
the dynamic aspects are represented by behavioral diagrams of
UML such as the interaction diagrams, state diagrams, and Activity diagram models the workflows and
activity diagrams. In the following we are trying to mention processes of business or System.
about different diagram very shortly. All diagrams grouped into
two groups 1st one is Structural diagrams and another one is Use Case and Activity diagram will be described later in
Behavioral diagrams. more detail.

a) Structural Diagrams B. Facts of Uses of Use Case Diagram
Class Diagram Use case diagrams represent the interaction between object
of system and the organizational units. But the Use case
Class diagram is most important and the main part of diagrams do not represent the process flow of a business. We
object-oriented systems. A class diagram shows a set can get the concept of activity and actors of those activity but
of classes together with their relationships. we can‟t get any sequence of activity and the condition of those
Component Diagram. activity. The use case model defines the relationships and
characteristics between business activities and participants
The component diagram mainly for repressing the outside the focused business, for example e.g. Patients and
relationship and dependencies among software Private transport, medicine suppliers etc. The object model
components. A component diagram includes focuses on the internal business processes object systems
Component, Component package. includes organization units, work units, workers and entities.
Deployment Diagram C. 4.3 Facts of Uses of Activity Diagram
The Deployment diagram of UML representation the Activity diagram of UML is one of the most used and
configuration of run time processing nodes and the useful diagram modeling the dynamic views / aspects of a
components that live on them. business process or system. It is important for modeling
processes and workflows. Activity diagrams can be applied as
b) Behavioral Diagrams business modeling method [17, 18]. Also Activity diagram can
Use Case Diagram. describe the sequence and branching of different business
processes in an organization with organizational unites very
The Use case diagram of UML represents the all Use clearly. It can describe components like activities, relationships
Case/function of a system and interaction of user with for control flow, logical connectors etc. In Activity diagram
the system. It represents the system from the user swimlines represent the organizational unites and grouped the
perspective. The elements of this diagram are: Actor, components. The UML activity diagrams have also strengths
Use-case, Association, Generalization. and weaknesses. The activity diagram support parallel
Interaction Diagram. behavior, but their great disadvantage is that they do not make
clear the links between activities and objects. However Activity
Interaction diagrams describe how a set of objects diagram is useful when we work with the following situations:
collaborates in some behavior. Interaction diagram is
two types, Sequence diagram, Collaboration diagram

129 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

D. Modeling an operation In use case diagram there is difficult to trace the
We can use Activity diagrams to model an operation of a iteration and sequence of execution
business process which we can mention in use case or a class. The properties of different item is not possible to
It is possible to model all operation with the relationship within display in diagrams
themselves Modeling a workflow
Activity diagram is the best way for modeling workflows V. COMPARISON BETWEEN EPC AND UML
across organization that involve many actors or business MODELING LANGUAGE
organizations. We can focus on activities and the actors which From the general point of view the both EPC and UML are
collaborate with the system. very useful for modeling business process. EPC and UML are
efficient in different aspects. In General we can say that the
E. Why UML EPC is a Process oriented modeling language and the UML is a
In above paragraph already we have described some Object Oriented Modeling Language. A more comprehensive
important features of UML which are indicates, why we use discussion can be found in [18]. The EPC is relevant for
UML for modeling. Also there are some other issues we business process modeling. Each of the UML diagram has
mentioned here. some aspects which are also relevant for business process
modeling. However, each diagram type has a certain focus for
UML is most widely used technical OO designed modeling business processes and also making information
UML is suitable for business modeling system.
UML is more structured
UML is more descriptive A. Relationship between EPC and Use case diagram:
UML have more graphical diagram A use case diagram has more interactions with a software
UML is more communicative for programmer. system and less to business-related functionality. There are
UML is easy to avoid the deadlock. three types of connections between an use case and an EPC.
UML can show the actor responsible to execute the 1. A use case diagram can specify a function of an EPC.
specific function. From an EPC diagram an EPC function can be
UML language is independent represent by several Use Cases e.g. check status,
UML is formal handle big project. check product catalogue, edit text and print text.
Effective for modeling large and complex software 2. In EPC a process can describe in underlying
business process. sequence, also it can be describe it in verbally, in a
Simple to learn UML support more tools more precisely way.
3. In EPC, involved people or objects involved with
F. Advantage of UML process can be interpreted as an organizational unit.
UML is more structured and has better attribute But In use case diagram it interpreted as actor. In all
UML has the flow of activities with different levels cases, we can use the same roles in both diagrams.
of attributes 4. The EPC diagram means a sequence of all process. In
Support for development process use case diagram we can represent all use case
function and actors but there is no process sequence.
Simplicity and its state clearly where the relationship
5. In EPC there is a process sequence so there is start
lies
and end event. But in use case diagram there is no
Top down methods are available
sequence so no start and end position.
Support for the development process
UML is easy to redraw any time and time saving for In EPC diagram the organization unit mentioned the using
re-engineering elapse shape tool in different place the organizational unit
Easy to understand for the both programmers and mentioned with different corresponding process. Here the
business people organizational unit is scattered. In use case diagram the all
organizational units are represented by actor and all use case
Easy to translate into code function are connected to actor .For example refer patient
No mathematical representation linked with two actor general practitioners and patient
UML is best suitable for large cases
Transport Male
G. Disadvantage of UML patient nurse
Necessary to specify system requirement
Lacking of showing logical relationship between
different functions and process
Use Case
Lacking of showing data flows and workflows.
Modeling dependency of two parallel task
Loop of care cycle until patient recover Figure 3: Organizational units in EPC and Use case diagram
Actor1 Actor2

130 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

B. Relationship between EPC and Activity diagram: necessary to be careful about exactness or ambiguity. From the
The Activity diagram and EPC are mainly serves the same analysis of elements interacting with process flow in EPC
purpose. So no need to use the EPC and UML activity diagram diagram may arise ambiguities. The following ambiguities are:
for modeling a same business processes. Also it is possible to Conjunctions of start event
translate EPCs diagram into activity diagrams and vice versa.
Of course, if we translate an EPC diagram to activity diagram In EPC modeling the start event is consider a node
then we can loss some information in an EPC. Because, the without input edge and similarly end event is consider
EPC cover a wider range of information than activity diagrams. the node without output edge. But ambiguity exists if
Activity diagram of UML does not consider external flows for start events found in the middle of the process.
modeling object. Deadlock and loops
C. Comparison between EPC and UML Activity Diagram It is very easy to analyze a simple graph in EPC but a
If we compare the EPC and UML Activity Diagram for tool is needed to analyze the complex graph. A
modeling business processes, there are some different aspects deadlock mean when a event start after the process
by which we can view the correspondences and differences runs and after some time the process is trapped and
between these two approaches. There are three following main unable to reach the end states. Deadlock may occur if
aspects for comparing the EPC and UML Activity diagram. logical connectors are mismatch; especially for
complex graph, there are different interpretation of
1. Context logical connectors and where one connector is link
2. Exactness/Ambiguity with other connector.
3. Notation/Terminology
c) Notation/Terminology
a) Context In comparison with EPC and UML activity diagram we can
This aspect describes the development and uses context of see that their concepts are similar but their representing style
EPC or UML Activity Diagram. Both diagrams are use for using different notation and terminology. There are some
modeling business processes, but both have different notations; those are not equivalent to other diagram. In our
development contexts. The EPC and UML Activity diagram discussion we will try to show the differences of both notation
drive the different modeling approaches. There are two and terminology and show some comparison of symbol of one
approaches to model a system: diagram to another.
Process-oriented modeling. Both EPC and UML activity diagram are comparable with
some common terminology; on the contrary they have
The Processes inside the system are the main focus in differences between them in correspondence with notation to
process-oriented modeling. A process consists of visualize the processes and workflows. Here we will try to
sequences of events and its activities/function. Event translate from EPC to UML activity diagram in correspondence
triggers activities/functions. An event is the results of with notational comparison and vice versa on latter. In the
other functions. From the first events, it trigger to the following figure 4 we try to shows some notational/
whole process. Some logical operator expands the terminology comparison between EPC and UML activity
control flow. The EPC flows the process oriented diagram:
modeling for standard business process modeling. The
basic EPC model can be extended by introducing
information objects and organization units with the
process.
Object-oriented modeling.
On the other hand, the object inside the system is the
main focus in object-oriented modeling. There are
different interrelated objects inside a system. These
independent objects communicate to each other by
exchanging messages. Each object has properties and
exchanges messages through operations. The UML
Activity diagram the object-oriented modeling
approach for developing IT solutions such as enterprise
information system. Figure 4: Comparison view of EPC and activity diagram

b) Exactness/Ambiguity We try to categorize the differences between both EPC and
For business process modeling there would be ambiguity UML diagram in connection with notation. These are as
both in EPC or activity diagram. For example there is a follows:
possibility of blocking for implicit decision. So that, for
designing diagram with EPC or UML activity diagram it is

131 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Functions and Activity/ Action states
The functions and activity/ action state in both EPC
and UML activity diagram activity diagrams states the
active elements that represent the organizational unit of
activity diagram and actors in use case diagram with
respect to the process.
F1 F1
Events
In EPC diagram an event is a result of the other
function that triggers to function. Activity diagram are
based on the state diagram and there is no relevant
events in activity diagram. F2 F3 F2 F3

Start and End State/Event Figure 6: Logical connector in EPC and Activity diagram
The EPC diagram start by an event with no input and Organizational unit and swim lane
end by an event with no output. But in activity diagram
start using a black circle and end by black circle with In EPC diagram organizational unit are attached with
another circle border. function which is responsible for the relevant business
task. On the other hand in activity diagram, use of
swim lane differentiate the activities belong to their
respective organizational unit. Compared with EPC the
activity diagram has some problem. Using the
swimlanes we can represent the organizational unit.
But some time the organization units are not enough
Figure 5: Starting and ending of EPC and activity diagram
for representing all organizational relationship. Say for
Data/Information Unit example, responsible for, provides support for, must be
informed about result of, and must approve of.
In EPC diagram we can mention data/ information unit
using by rectangle but an activity or use case diagram In EPC diagram of patient Admission module the
there is no tools for representing external data. organization unit mentioned the using elapse shape tool
in different place the organizational unit mentioned
Control flow and transaction with different corresponding process . For example
In comparison with the control flow of EPC and transport patient by male nurse, fill form by physician
transaction in activity diagram, both are similar. In update ward book by nurse.
EPC modeling control flow is used to representing the In activity diagram all organizational units are
process chain that is; one event triggers a business organized in some swimlanes and the different function
function which is the result of other function. In has been placed within different swimlanes .for
activity diagram transitions show the change of state example forwarded Admission document, update word
over time and these are based on state diagram. book. Enter in pc etc has been placed in nurse. When a
Logical connector function done by two or more organizational units it is
too difficult to manage in swimlanes in Activity
For considering the EPC modeling logical connector diagram.
permit for splitting control flow, on the other hand
activity diagram used transition to perform this Iteration
splitting. Both the diagram used similar branch/ merge In EPC diagram there is no iterative notation, on the
for splitting control flow and taking decision. For other hand activity diagram support iteration notation.
taking decision EPC used logical „XOR‟ connector and
for the same operation activity diagram used decision
diamond. For parallel activity EPC use Logical „AND‟ VI. CONCLUSION
connector while in activity diagram use The purposes of business process modeling to make an
synchronization bar. The main difference between EPC overall view and workflow of a business. For modeling the
and UML activity diagram is that there is no notation hospital case we have discussed about EPC and UML and we
to denote „OR‟ operation in activity diagram like EPC have got lots of advantage and disadvantage of language. Also
diagram. at the time of making modeling diagram of any business or
organization, we found some difficulties and ambiguity. We
have tried to discuss about those language and the modeling
process which is helpful to make a concept of business model
as well as modeling languages. Those concepts will help to
continue research the business process modeling and Model
any business process.\

132 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

REFERENCE Saarbrükken, March 1998. (http://www.tu-
chemnitz.de/wirtschaft/wi2/home/loos/iwih144.pdf)
[1] Balbir Barn Feb 2007, “Business Process Modeling”
[22] Rittgen, P.: EMC - A Modeling Method for Developing Web-based
[2] http://www.jisc.ac.uk/media/documents/programmes/eframework/proces Applications. International Conference of the International Resources
s-modelling_balbir_barn.pdf Management Association (IRMA) 2000, Anchorage, Alaska, USA, May
[3] KELLER, G.; NÜTTGENS, M.; SCHEER, A.-W. (1992): Semantische 21 - 24, 2000
rozeßmodellierung auf der Grundlage ''Ereignisgesteuerter Prozeßketten [23] Ferdian, “A Comparison of Event-driven Process Chains and UML
(EPK)'', Veröffentlichungen des Instituts für Wirtschaftsinformatik, Activity Diagram for Denoting Business Processes”, Technische
Heft 89, Saarbrücken, Universität Hamburg-Harburg Arbeitsbereich Softwaresysteme, April
[4] URL: http://www.iwi.uni-sb.de/public/iwihefte/ heft089.zip. 1st, 2001
[5] NÜTTGENS, M. (1997): Event-driven Process Chain (EPK) - Some [24] Van der Aalst W. (1999), “Formalization and verification of Event-
Links and Selected Publications, URL: http://www.iwi.uni- driven Process Chains”, Information and Software Technology 41, pp.
sb.de/nuettgens/EPK/epk.htm. 639-650.
[6] Keller, G.; Meinhardt, S.: SAP R/3-Analyzer: Optimierung von
Geschäftsprozessen auf der Basis des R/3-Referenzmodells, Walldorf, AUTHORS PROFILE
1994.
Md. Rashedul Islam
[7] Scheer A-W, (2000). ARIS: Business Process Modelling, 3rd Edition.
Berlin, Springer. He has completed Bachelor of Science (Hons.) in Computer Science and
Engineering at University of Rajshahi, Bangladesh on 2006, Now he is
[8] Scheer A-W, (2002), “ARIS – From the vision to practical process studying Master‟s of Informatics under School of Business and
control”, in Business Process Excellence, Berlin, Springer, pp. 1-14. Informatics, Högskolan i Borås, Sweden. This M.Sc is mostly finish. He
[9] IDS Scheer, (2003) ARIS Toolset, version 6.2 is a Senior Lecturer, Dept of CSE, Leading University, Bangladesh, also
[10] Vernadat F. (2002), “Enterprise modeling and integration (EMI): have 5 years IT development experience; He has two Journal
Current status and research perspectives”, Annual Reviews in Control publications and five Conference publications. His current research
26, pp. 15-25. interest: Parallel Programming, Signal & Speech Processing,
Management Information, Information system planning, Software
[11] Gartner Group (2002), Gartner Group Report. Available at Engineering.
http://www.gartner.com.
[12] Gulledge T., and Sommer R. (2002), “Business process management:
public sector implications”, Business Process Management Journal 8, Md. Rofiqul Islam
pp. 364-376. He has completed Bachelor of Science (Hons.) in Computer Engineering
[13] A.-W. Scheer. Business Process Engineering. Reference Models for at American International University of Bangladesh 2006, Now he is
Industrial Enterprises. Springer-Verlag, 1995. studying Master‟s of Informatics under School of Business and
Informatics, Högskolan i Borås, Sweden. This M.Sc is mostly finish. He
[14] Unified Modeling Language, http://www-
has one Journal publication. His current research interest: Parallel
01.ibm.com/software/rational/uml/ at 23-10-2010
Programming, Information system planning, Software Engineering.
[15] G. Booch, J. Rumbaugh, I. Jacobson. The Unified Modeling Language
Reference Manual. Addison Wesley, 1999.
Md. Shariful Alam
[16] Ambler, S. W.: What's Missing from the UML? Techniques that can
help model effective business applications, in: Object Magazine, 7 He has completed BSc(Eng.) in Computer Science and Engineering from
(1997) 8. Chittagong University of Engineering and Technology, Chittagong,
Bangladesh on 2008, Now he is studying Master‟s of Informatics under
[17] [BRJ99b] G. Booch, J. Rumbaugh, I. Jacobson. The Unified Modeling
School of Business and Informatics, Högskolan i Borås, Sweden. This
Language User Guide. Addison Wesley, 1999.
M.Sc is mostly finish. He has almost 4 years professional experience in
[18] P. Loos, T. Allweyer. Process Orientation and Object Orientation – An softwere development field. His current research interest: Object
Approach for Integrating UML and Event-Driven Process Chain (EPC). Oriented Techonology, Image Processing, Software Engineering,
Publication of the Institut für Wirtschaftsinformatik, Paper 144, System Development Philosophy, Information System and Business
Saarbrücken, 1998 (http://www.iwi.uni-sb.de/iwi-hefte/iwih144.ps). Process.
[19] Fowler, M.; Scott, K.: UML Distilled – A Brief Guide to the Standard Md. Shafiul Azam
Object Modeling Language. 2nd ed. Reading, MA 1999.
He has completed Bachelor of Science (Hons.) in Computer Science and
[20] Paech, B.: On the Role of Activity Diagrams in UML, in: P.-A. Muller, Engineering at University of Rajshahi, Bangladesh on 2006 and M.Sc in
J. Bezivin, (eds.), Proceedings of the Workshop <<UML>>‟98, Beyond Computer Science and Engineering at University of Rajshahi,
the Notation (Mulhouse, June 3-4, 1998), pp 245-250.2. Bangladesh on 2008. He is a Lecturer, Dept of CSE, Science and
[21] Loos, P; Allweyer, T.: Process Orientation and Object-Orientation - An Technology University, Pabna, Bangladesh. He has two Conference
Approach for Integrating UML and Event-Driven Process Chains (EPC). publications. His current research interest: Image Processing,
Publication of the Institut für Wirtschaftsinformatik, Paper 144, Information system planning, Software Engineering.

133 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

FACIAL TRACKING USING RADIAL
BASIS FUNCTION
P.Mayilvahanan, Dr.S.Purushothaman Dr.A.Jothi
Research scholar, Principal, Sun College of Dean,
Dept. of MCA, Engineering & Technology, School of Computing
Vel's University, Kanyakumari – 629902, Sciences, Vel’s University,
Pallavaram, Chennai, India Tamil Nadu, India, Email: Pallavaram, Chennai, India
<dr.s.purushothaman@gmail.
com

ABSTRACT--This paper implements facial tracking the template, by projecting it around the detected
using Radial basis function neural network (RBF). positions of the target and considering its overlap
There is no unique method that claims perfect facial with the segmented local object. The tracking
tracking in video transfer. The local features of a
results show good performance, when the camera
frame are segmented. A ratio is found based on a
criteria and output of RBF is used for transferring the
moves towards the object. Yan Tong et al. [3]
necessary information of the frame from one system to developed a general framework for region
another system. A decision approach, with a threshold, tracking which includes models for image
is used to detect if there is any change in the local changes due to motion, illumination and partial
object of the successive frames. The accuracy of the occlusion. They used a cascaded parametric
result depends upon the number of centers. The motion model and a small set of basis images, to
performance of the algorithm in reconstructing the account for shading changes, which will be
tracked object is about 96.5% and similar to the solved in a robust estimation framework, in order
performance of back propagation algorithm (BPA), in
to handle small partial occlusion. Gleicher [4]
terms of reduced time and quality of reconstruction.
introduced difference decomposition, to solve
Index Terms- Radial basis function (RBF), Back- the registration problem in tracking, where the
propagation algorithm (BPA); Watershed algorithm; difference would be linear combination of a set
Motion Estimation. of basis vectors. Sclaroff and Isidoro [5] used
this idea for template registration in region-based
1. INTRODUCTION non-rigid tracking, where the non-rigid
deformation was represented, in terms of
In specific applications like video- eigenvectors of a finite element method.
conferencing, news telecast, most of the image Photometric variation is considered; and a
area is covered by a human face. Low bit-rate modified Delaunay refinement algorithm is used
video transmission is required by using 3D head to construct a consistent triangular mesh for the
models. Tracking algorithms are available to region of the tracked object.
track head in video sequence. There is no Nguyen and Worring [6] made their
complete automatic system available for contribution, by introducing a contour tracking
extracting head model from the video. If three method, incorporating static segmentation by the
dimension head models can be extracted from watershed algorithm. Their method utilized kinds
the first frame (or a first few frames) in a video of edge maps from motion (optic flow), intensity
sequence then it will become possible to build (watershed) and prediction (contour warping), to
extremely low bit rate video coding systems for update the object contour. It was claimed that
communicating head and shoulder scenes. Head this method yielded accurate and robust results.
models can be used for synthesizing views and The idea of “active blob” by Nguyen
facial expressions, animating virtual characters. and Worring [6] discusses the non-rigid
Shi and Tomasi [1] put forward the deformation. The Delaunay triangulation of
criterion of “good features” by its texture and computer graphics is used to generate some mesh
used it in affine feature tracking. Parry et. al [2] of the object region [5].
introduced a region-based (formed by
segmentation) tracking method, mainly updating

134 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

In general, the following procedure is B. Methods
adopted for face tracking algorithm [7]
The Radial basis function (RBF) is a
a. Wait for a face(s) to appear in the frame supervised artificial neural network method [8].
b. Enter initialization mode (wait for a face(s) to The concept of distance measure is used to
appear for a predefined amount of time, to avoid associate the input and output pattern values, eq
paying attention to people who just happen to (1). Radial Basis Functions is capable of
pass by) producing approximations to an unknown
b. Enter tracking mode, choose the closest face. function ‘f’ from a set of input data abscissa. The
c. Track the face until it leaves the frame. To approximation is produced by passing an input
avoid losing track of the face due to minor head point through a set of basis functions, each of
movements, leave tracking mode only when the which contains one of the RBF centers,
tracked face disappears for a predefined amount multiplying the result of each function by a
of time. coefficient and then summing them linearly. For
d. Go to a. each function ‘t’, the approximation to this
function is essentially stored in the coefficients
This paper proposes a region-based and centers of the RBF. These parameters are in
method for motion estimation undergoing object no way unique, since for each function ‘t’ is
tracking. Tracking is performed by means of approximated, many combinations of parameter
motion segmentation. The proposed method fully values exist. RBFs have the following
mathematical representation:
utilizes information of temporal motion and
spatial luminance. Computation of dominant N −1
motion of the tracked object is done by a robust F(x) = ∑ c i Φ(|| x − R i ||) (1)
iterative weights least square (IWLS) method. i =0
Static segmentation is incorporated to modify
this prediction, where the warping error of each
watershed segment and its rate of overlapping where
with warped template are utilized, to help
R is a vector containing the centers of the
classification of some possible watershed
RBF, and
segments near the object border.
The following procedure is used to φ is the basis function or activation function
implement RBF for facial tracking. of the network. The implementation of RBF for
• Read frame1 facial tracking is as follows:
• Take a portion of frame 1(eye/nose/lip etc.)
• Apply watershed segmentation
• Find mean of the segmented image Step 1: Apply Radial Basis Function.
• Train /test using RBF No. of Input = width of the facial parameter
• During testing , get the output of RBF in number of pixels
• Accordingly display image in system 2 No. of Patterns = No. of frames under
implementation
II MATERIALS AND METHODS No. of centres = No. of patterns
A. Materials Calculate RBF as
The concept of ANN, with supervised
algorithm for computing the affine RBF = exp (-X)
transformation-taking place in the current frame, Calculate
with respect to previous frame. This is achieved,
when there is a significant change in the output G = RBF
of the neural network, which indicates the A = GT * G
change in position of the object in the current
Calculate
frame. To detect the change in position of the
object, the network has to be trained in advance B = A-1
under supervised mode.
Calculate
E = B * GT

135 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Step 2: Calculate the Final Weight. uses final weights obtained during training for
F=E*D updating the frames in the receiving system.

Step 3: Store the Final Weights in a File. Here, fx , fy and ft are the partial
The final updated weights are saved for derivatives of brightness function with respect to
testing the video transfer. x, y and t, the function is chosen as the Nesi
and Magnolfi [8]; and σ is the scale parameter.
III SCHEMATIC DIAGRAM OF FACIAL To solve the problem, there are two different
TRACKING ways to find robustly the motion parameters: one
is gradient-based, like the SOR method in [3];
another is least squares-based, such as (IWLS)
method. The algorithm begins, by constructing
the Gaussian pyramid (three levels are set up).
When the estimated parameters are interpolated
into the next level, they are used; to warp
(realized by bilinear interpolation) the last frame
to the current frame. In the current level, only the
changes are estimated in the iterative update
scheme
(a) Training In static segmentation, the watershed
algorithm of mathematical morphology is a
powerful method [4]. Early watershed algorithms
are developed, to process digital elevation
models, and are based on local neighborhood
operations on square grids. Some approaches use
“immersion simulations“to identify watershed
segments, by flooding the image, with water
starting at intensity minima. Improved gradient
methods are devised, to overcome plateaus and
square pixel grids [10]. The former method is
used. A severe drawback, to the computation of
watershed algorithm, is over-segmentation.
Normally watershed merging is performed, along
with the watershed generation. Over-
segmentation is welcome; so, during tracking,
the merging process is omitted, which saves
some computational costs. Figure 2 shows
procedure for watershed segmentation.

IV TEMPLATE WARPING AND REGION
ANALYSIS

Once the motion parameters have been
(b)Testing computed, warp the object template from the last
frame to the current frame. Then the warped
Figure. 1 Procedure for facial tracking template is used to determine, which watershed
segments enter the template according, to the
Figure 1a shows the training procedure following measure: Given that the number of
for the RBF. Frames are extracted from the pixels belonging to the warped template in the
video. Watershed segmentation is applied on the number of all pixels in Ri is Ci, a ratio ri is
local and global objects in the face. The mean computed, as given in eq(2):
values of the segmented image are used for
training the RBF. The final weights are stored in ri =CPi / Ci (2)
a file. Figure 1b shows the procedure for
updating the frames in the receiving system. It

136 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

Based on this measure, the Human face motion is complex with
classification problem of each sub-region is rigid and non-rigid movements; hence the idea in
discussed in the following cases: [3] adopted, using a modified affine model, to
1) When ri > r0, then, classify ri as part of the describe the local motion of facial features
final object template; (mouth, eyes and eyebrows) and a planar
2) When r0 ≥ ri ≥ r1 (here r1 = 0.4), another projective transform to model the head motion.
measure as MAE (Mean Absolute Error) of The IWLS method is used to estimate these
difference between the warped frame and the motion parameters.
current frame is taken into account, eq (3)

V. EXPERIMENT RESULTS

The project is implemented using
Matlab 7. The time taken for processing each
frame is at an average of 1.4 seconds. This
includes segmentation and processing with
neural network. The topology of the RBF is
width of the FAP x number of patternsx1. The
total number of frames in the video is 91. The
peak signal to noise ratio (PSNR) for all the
reconstructed face in the received system is
shown in Figure 3. For the experiment, only 8
frames (Figure 4) are considered, which show
significant changes in the lip movements.
39

37
peak signal to noise ratio

35
Figure 2 Flow chart for watershed algorithm
34

M i = ∑ | f ( x, t + 1) − f w ( x, t ) | / Ci
33

(3) 32

where fw (x , t ) the warped image of f(x ,t), 31
using the estimated dominant motion parameters; 0 10 20 30 40 50
Frame numbers updated
60 70 80

If the warped error Mi of Ri is smaller enough Figure.3 Peak Signal to noice ratio
(less than a of f (x , t ) using given threshold, for
instance, 10), Ri is still regarded as part of the VI. CONCLUSION
updated template; otherwise, exclude Ri out of
the object region. In this paper, an RBF based approach is
3) When ri < r1, ri will not be included in the proposed for motion estimation, undergoing
updated template. facial tracking. The lip movements are mainly
focused in this work. The template warping, by
When people make facial expression watershed segmentation and ANN for quick
movements, especially behaving emotionally, decision of frame updation, is implemented.
(mainly, six universal facial expressions are to be Applications of this method in facial expression
discussed, i.e. disgust, sadness, happiness, fear, tracking can be expressed for other parts of face.
anger and surprise), in most of cases head motion
is accompanied. The procedure is divided into REFERENCES
two steps: 1) Head tracking is realized first, then [1] Shi J. and Tomasi C, “Good features to
the estimated motion is used to stabilize the face track”. In Proc. Computer Vision and Pattern
region; 2) The local motion of each facial feature Recognition, 1994.
is estimated relative to the stabilized face. [2] Parry et. al, “Region Template Correlation
for FLIR Target Tracking”, British Machine
Vision Conference’96.

137 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, March 2011

[3] Yan Tong, YangWang, Zhiwei Zhu, Qiang Ji,
Robust facial feature tracking under varying face
pose and facial expression, Pattern Recognition
40 (2007) 3195 – 3208.
[4] Gleicher M, “Projective registration with
difference decomposition”, IEEE CVPR’97,
pp331-337, 1997.
[5] Sclaroff S. and Isidoro J., “Active blobs”,
ICCV’98. Frame no:5 Frame no:16
[6] Nguyen, Worring M., “Multi-feature object
tracking using a model-free approach”, IEEE
CVPR, pp 145 –150, 2000.
[7] Anton Podolsky and Valery
Frolov, Face tracking,
www.cs.bgu.ac.il/~orlovm/
teaching/saya/.../saya-tracking-report.pdf Frame no:24 Frame no:37
[8] P. Nesi and R. Magnolfi, Tracking and
Synthesizing Facial Motions with Dynamic
Contours Real-Time Imaging 2, 67–79 (1996)
[9] Vincent L, Soille, “Watersheds in digital
spaces: an efficient algorithm based on
immersion simulations”, IEEE T-PAMI, 13(6):
583-589, 1991.
[10] Gauch J, „Image segmentation and analysis Frame no:46 Frame no:57
via multi-scale gradient watershed hierarchies“,
IEEE T-IP, 8(1): 69-79, 1999.

Frame no:70 Frame no:91

Figure 4 Experimental results relating to lip movements

138 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011

Performance Comparison of Speaker Identification
using circular DFT and WHT Sectors
Dr. H B Kekre1, Vaishali Kulkarni2,
Indraneal Balasubramanian3, Abhimanyu Gehlot4, Rasik Srinath5
1
Senior Professor, Computer Dept., MPSTME, NMIMS University.
hbkekre@yahoo.com
2
Associate Professor, EXTC Dept., MPSTME, NMIMS University.
Vaishalikulkarni6@yahoo.com
3, 4, 5
students, B-Tech EXTC, MPSTME, NMIMS University.
indraneal89@gmail.com, abhimanyu13090@gmail.com, rasik90@gmail.com

Abstract— In this paper we aim to provide a unique approach to identification using power distribution in the frequency domain
text dependent speaker identification using transform techniques [11], [12]. We have also proposed speaker recognition using
such as DFT (Discrete Fourier Transform) and WHT (Walsh vector quantization in time domain by using LBG (Linde Buzo
Hadamard Transform). In the first method, the feature vectors Gray), KFCG (Kekre’s Fast Codebook Generation) and KMCG
are extracted by dividing the complex DFT spectrum into (Kekre’s Median Codebook Generation) algorithms [13 – 15]
circular sectors and then taking the weighted density count of the and in transform domain using DFT (Discrete Fourier
number of points in each of these sectors. In the second method, Transform), DCT (Discrete Cosine Transform) and DST
the feature vectors are extracted by dividing the WHT spectrum (Discrete Sine Transform) [16].
into circular sectors and then again taking the weighted density
count of the number of points in each of these sectors. Further, The concept of sectorization has been used for (CBIR)
comparison of the two transforms shows that the accuracy content based image retrieval. [17] – [21]. We have proposed
obtained for DFT is more (80%) than that obtained for WHT speaker identification using circular DFT sectors [22]. In this
(66%). paper, we propose speaker identification using WHT (Walsh
Hadamard Transform), and also compare the results with DFT
Keywords - Speaker identification; Circular Sectors; weighted sectors. In Fig. 1, we can see how a basic speaker identification
density; Euclidean distance system operates. A number of speech samples are collected
from a variety of speakers, and then their features are extracted
I. INTRODUCTION and stored as reference models in a database. When a speaker is
Human speech conveys an abundance of information, from to be identified, the features of his speech are extracted and
the language and gender to the identity of the person speaking. compared with all of the reference speaker models. The
The purpose of a speaker recognition system is thus to extract reference model which gives the minimum Euclidean distance
the unique characteristics of a speech signal that identify a with the feature vector of the person to be identified is the
particular speaker [1 - 4]. Speaker recognition systems are maximum likelihood model and is declared as the person
usually classified into two subdivisions, speaker identification identified.
and speaker verification [2 – 5]. Speaker identification (also
known as closed set identification) is a 1: N matching process II.
where the identity of a person must be determined from a set of
known speakers [7]. Speaker verification (also known as open III.
set identification) serves to establish whether the speaker is
who he claims to be [8]. Speaker identification can be further IV.
classified into text-dependent and text-independent systems. In
a text dependent system, the system knows what utterances to V.
expect from the speaker. However, in a text-independent
system, no assumptions about the text can be made, and the
system must be more flexible than a text dependent system [4, VI.
5, and 8].
VII. EASE OF USE
Speaker recognition systems find use in a multitude of
applications today including automated call processing in A. Selecting a Template (Heading 2)
telephone networks as well as query systems such as stock FF
information, weather reports etc. However, difficulties in wide
deployment of such systems are a practical limitation that is yet
to be overcome [2, 6, 7, 9, and 10]. We have proposed speaker Figure 1. Speaker Identification System

139 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
II. SECTORIZATION OF THE COMPLEX TRANSFORM PLANES
The speech signal has amplitude range from -1 to +1. It is
A. Discrete Fourier Transform(DFT) first converted into positive values by adding +1 to all the
The DFT transforms time or space based data into sample values. Thus the amplitude range of the speech signal is
frequency-based data. The DFT allows you to efficiently now from 0 to 2. For sectorization two methods are used,
estimate component frequencies in data from a discrete set of which are described below:
values sampled at a fixed rate [23, 24]. If the speech signal is
represented by y (t), then the DFT of the time series or A. DFT Sectorization
samples y0, y1,y2, …..yN-1 is defined as given by (1): The algorithm for DFT sectorization is given below:
1. The DFT of the speech signal is computed. Since the DFT
-2jπkn/N
Yk = ne
is symmetrical, only half of the number of points in the
DFT is considered while drawing the complex DFT plane
(1)
(i.e. Yreal vs. Yimag).
Where yn=ys (nΔt); k= 0, 1, 2…, N-1.
Δt is the sampling interval. 2. Also the first point in DFT is a real number, so it is
considered separately while taking feature vectors. So the
B. Walsh Hadamard Transform complex plane is only from (2, N/2), where N is the
The Walsh transform or Walsh–Hadamard transform is a number of points in DFT. Fig. 2 shows the original speech
non-sinusoidal, orthogonal transformation technique that signal and its complex DFT plane for one of the samples
decomposes a signal into a set of basis functions. These basis in the database.
functions are Walsh functions, which are rectangular or square
waves with values of +1 or –1. The Walsh–Hadamard 3. For dividing the complex plane into sectors, the
transform returns sequency values. Sequency is a more magnitude of the DFT is considered as the radius of the
generalized notion of frequency and is defined as one half of circular sector as in (3):
the average number of zero-crossings per unit time interval.
Each Walsh function has a unique sequency value. You can Radius (R) = abs (sqrt ((Yreal)2+(Yimag)2)) (3)
use the returned sequency values to estimate the signal
frequencies in the original signal. The Walsh–Hadamard 4. Table I shows the range of the radius taken for dividing
transform is used in a number of applications, such as image the DFT plane into circular sectors.
processing, speech processing, filtering, and power spectrum
analysis. It is very useful for reducing bandwidth storage 1

requirements and spread-spectrum analysis [25]. Like the FFT, 0.5
the Walsh–Hadamard transform has a fast version, the fast
Amplitude

Walsh–Hadamard transform (fwht). Compared to the FFT, 0

the FWHT requires less storage space and is faster to calculate
-0.5
because it uses only real additions and subtractions, while the
FFT requires complex values. The FWHT is able to represent -1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
signals with sharp discontinuities more accurately using fewer No. of samples 4
x 10

coefficients than the FFT. FWHTh is a divide and conquer 400

algorithm that recursively breaks down a WHT of size N into
300
two smaller WHTs of size N / 2. This implementation follows
the recursive definition of the Hadamard matrix HN 200

given by (2): 100
Ximag

-100

(2)
The normalization factors for each stage may be -200

grouped together or even omitted. The Sequency ordered, also -300

known as Walsh ordered, fast Walsh–Hadamard transform,
-400
FWHTw, is obtained by computing the FWHTh as above, and -400 -300 -200 -100 0
Xreal
100 200 300 400

then rearranging the outputs.
The rest of the paper is organized as follows: Section II
explains the sectorization process, Section III explains the Figure 2. Speech signal and its complex DFT plane
feature extraction using the density of the samples in each of
the sectors, Section IV deals with Feature Matching, and results 5. The maximum range of the radius for forming the sectors
are explained in Section V and the conclusion in section VI. was found by experimenting on the different samples in

Identify applicable sponsor/s here. (sponsors)

140 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
the database. Various combinations of the range were
tried and the values given in Table I was found to be 300

satisfactory. Fig. 3 shows the seven sectors formed for the
complex plane shown in Fig. 2. Different colours have 200

been used to show the different sectors.

6. The seven circular sectors were further divided into four 100

quadrants each as given by Table II. Thus we get 28
sectors for each of the samples. Fig. 4 shows the 28 0

sectors formed for the sample shown in Fig. 2.
-100

TABLE I. RADIUS RANGE OF THE CIRCULAR SECTORS
-200
Sr. Radius range Sector Weighing
No. factor
1 0≤R≤4 Sector1 2/256 -300
2 4≤R≤8 Sector2 6/256 -300 -200 -100 0 100 200 300

3 8≤R≤16 Sector3 12/256
4 16≤R32 Sector4 24/256
5 32≤R≤64 Sector5 48/256 Figure 4. Sectorization of DFT plane into 28 sectors for the speech
6 64≤R≤128 Sector6 96/256 sample shown in Fig. 2
7 128≤R≤256 Sector7 192/256

1. The WHT of the speech signal is taken using FWHT
250
(Fast Walsh Hadamard Transform).
225

200 2. The WHT can be represented as (C0, S0, C1, S1, C2,
175

150
S2, …….., CN-1, SN-1), C represents Cal term and S
125 represents Sal term.
100

75
3. The Walsh transform matrix is real but by
50

25
multiplying all Sal Components by j it can be made
0 complex. The first term i.e. C0 represents dc value. So
-25

-50
the complex plane is considered by combining S0
-75 with C1, S1 with C2 and so on. In this case SN-1 will be
-100 left out. Thus C0 and SN-1 are considered separately.
-125

-150

-175
4. The complex Walsh transform is then divided into
-200 circular sectors as shown by (4). Again the radial
-225
sectors are formed using the radius as shown in Table
-250
-250 -225 -200 -175 -150 -125 -100 -75 -50 -25 0 25 50 75 100 125 150 175 200 225 250
I.

Figure 3. Circular Sectors of the complex DFT plane of the speech Radius (R) = abs (sqrt ((Ycal)2+(Ysal)2)) (4)
sample shown in Fig. 2
5. The seven circular sectors were further divided into
TABLE II. DIVISION INTO FOUR QUADRANTS four quadrants as explained in (A) by using Table II.
Thus we get 28 sectors for each of the samples.
Sr. value Quadrant
No.
1 Xreal≥0 & Ximag≥0 1 (00 – 900 ) III. FEATURE VECTOR EXTRACTION
2 Xreal≤0 & Ximag≥0 2 (900 – 1800) For feature vector generation, the count of the number of
3 Xreal≤0 & Ximag≤0 3 (1800 – 2700) points in each of the sectors is first taken. Then feature vector
4 Xreal≥0 & Ximag≤0 4 (2700 – 3600) is calculated for each of the sectors according to (5).

Feature vector = ((count/n1)*weighing factor)*10000 (5)
B. WHT Sectorization
The algorithm for Walsh Sectorization is given below: For DFT, the first value i.e. dc component is accounted as in
(6). For WHT, C0 is accounted as given by (6) and SN-1 is
considered as given by (7). Overall there are eight components
in the feature vector for DFT (one per sector and first term).

141 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
Similarly, there are nine components in the feature vector for decreases. When the complex plane is further divided into 56
WHT (one per sector, first term and last term), when the seven sectors, there is a improvement in accuracy for less number of
circular sectors are considered. When 28 sectors are samples, but as the number of samples is increased
considered there are 29 components in the feature vector (one performance is similar as that with 28 sectors. Fig. 6 shows the
per sector and first term) for DFT and 30 components in the
feature vector (one per sector, first term and last term) for
WHT.

First term = sqrt (abs (first value of DFT/WHT)) (6)

Last term = sqrt (abs (Last value of FWHT)) (7)

IV. RESULTS

A. Database description
The speech samples used in this work are recorded using
Sound Forge 4.5. The sampling frequency is 8000 Hz (8 bit,
mono PCM samples). Table II shows the database description.
The samples are collected from different speakers. Samples are
taken from each speaker in two sessions so that training model
and testing data can be created. Twelve samples per speaker are
taken. The samples recorded in one session are kept in database
and the samples recorded in second session are used for testing.

TABLE III. DATABASE DESCRIPTION
Figure 5. Accuracy for DFT Sectorization
Parameter Sample characteristics
Language English
No. of Speakers 30
Speech type Read speech
Recording conditions Normal. (A silent room)
Sampling frequency 8000 Hz
Resolution 8 bps

B. Experimentation
This algorithm was tested for text dependent speaker
identification. Feature vectors for both the methods described
in section II were calculated as shown in section III. For
testing, the test sample is similarly processed and feature vector
is calculated. For recognition, the Euclidean distance between
the features of the test sample and the features of all the
samples stored in the database is computed. The sample in the
database for which the Euclidean distance is minimum, is
declared as the speaker recognized.

C. Accuracy of Identification
The accuracy of the identification system is calculated as
given by equation 5.

(5)

Fig. 5 shows the results obtained for DFT sectorization. As Figure 6. Accuracy for WHT Sectorization
seen from the results, when the complex DFT plane is divided
into seven sectors, the maximum accuracy is around 80% and results obtained for WHT sectorization. Here also we see that
decreases as the number of samples in the database is increased accuracy improves as the number of sectors is increased from
(64% for 30 samples). It can be seen that accuracy increases 7 to 28. But further division into 56 sectors does not give any
when the number of sectors into which the complex DFT plane advantage. Overall the results obtained for DFT are better than
is divided, is increased from 7 to 28. With 28 sectors, the those obtained for WHT.
maximum accuracy is 80% up to 20 samples after which it

142 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 9, No. 3, 2011
V. CONCLUSION [17] H B Kekre, Dhirendra Mishra, “Performance Comparison of Density
Distribution and Sector mean of sal and cal functions in Walsh
Speaker Identification using the concept of Sectorization has Transform Sectors as Feature Vectors for Image Retrieval ” ,
been proposed in this paper. The complex DFT and WHT International Journal of Image Processing ,Volume :4, Issue:3, 2010.
plane has been divided into circular sectors and feature vectors [18] H B Kekre, Dhirendra Mishra, “CBIR using Upper Six FFT Sectors of
Color Images for Feature Vector Generationl”, International Journal of
have been calculated using weighted density. Accuracy Engineering and Technology,Volume :2(2) ”, 2010.
increases when the 7 circular sectors are divided into 28 [19] H B Kekre, Dhirendra Mishra, “Performance Comparison of Four,
sectors for both the transform techniques. But there is no Eight & Twelve Walsh Transform Sectors Feature Vectors for Image
significant improvement when the complex plane is further Retrieval from Image Databases”, International Journal of Engineering
divided. The results also show that the performance of DFT is Science and Technology”, Volume :2(5) , 2010.
better than WHT. [20] H B Kekre, Dhirendra Mishra, “ Four Walsh Transform Sectors
Feature Vectors for Image Retrieval from Image Databases ” ,
International Journal of Computer Science and Information
Technologies”, Volume :1(2) , 2010.
REFERENCES [21] H B Kekre, Dhirendra Mishra, “Digital Image Search & Retrieval
using FFT Sectors of Color Images”, International Journal of Computer
[1] Lawrence Rabiner, Biing-Hwang Juang and B.Yegnanarayana, Science and Engineering”, Volume :2 , No.2, 2010.
“Fundamental of Speech Recognition”, Prentice-Hall, Englewood Cliffs, [22] H B Kekre, Vaishali Kulkarni, “Automatic Speaker Recognition using
2009. circular DFT Sector”, Interanational Conference and Workshop on
[2] S Furui, “50 years of progress in speech and speaker recognition Emerging Trends in Technology (ICWET 2011), 25-26 February, 2011.
research”, ECTI Transactions on Computer andInformation Technology, [23] Bergland, G. D. "A Guided Tour of the Fast Fourier Transform." IEEE
Vol. 1, No.2, November 2005. Spectrum 6, 41-52, July 1969
[3] D. A. Reynolds, “An overview of automatic speaker recognition [24] Walker, J. S. Fast Fourier Transform, 2nd ed. Boca Raton, FL: CRC
technology,” Proc. IEEE Int. Conf. Acoust., Speech,S on Speech and Press, 1996.
Audio Processing, Vol. 7, No. 1, January 1999. IEEE, New York, NY, [25] Terry Ritter, Walsh-Hadamard Transforms: A Literature Survey, Aug.
U.S.A 1996.
[4] S. Furui. Recent advances in speaker recognition. AVBPA97, pp 237--
251, 1997
[5] J. P. Campbell, ``Speaker recognition: A tutorial,'' Proceedings of the
IEEE, vol. 85, pp. 1437--1462, September 1997. AUTHORS PROFILE
[6] D. A. Reynolds, “Experimental evaluation of features for robust speaker
identification,” IEEE Trans. Speech Audio Process., vol. 2, no. 4, pp. Dr. H. B. Kekre has received B.E. (Hons.) in Telecomm. Engg. from Jabalpur
639–643, Oct. 1994. University in 1958, M.Tech (Industrial Electronics) from IIT Bombay in 1960,
[7] Tomi Kinnunen, Evgeny Karpov, and Pasi Fr¨anti, “Realtime Speaker M.S.Engg. (Electrical Engg.) from University of Ottawa in 1965 and Ph.D.
Identification”, ICSLP2004. (System Identification) from IIT Bombay in 1970. He
has worked Over 35 years as Faculty of Electrical
[8] F. Bimbot, J.-F. Bonastre, C. Fredouille, G. Gravier, I. Magrin- Engineering and then HOD Computer Science and Engg.
Chagnolleau, S. Meignier, T. Merlin, J. Ortega-García, D.Petrovska- at IIT Bombay. For last 13 years worked as a Professor in
Delacrétaz, and D. A. Reynolds, “A tutorial on text-independent speaker Department of Computer Engg. at Thadomal Shahani
verification,” EURASIP J. Appl. Signal Process., vol. 2004, no. 1, pp.
Engineering College, Mumbai. He is currently Senior
430–451, 2004.
Professor working with Mukesh Patel School of Technology Management and
[9] Marco Grimaldi and Fred Cummins, “Speaker Identification using Engineering, SVKM’s NMIMS University, Vile Parle(w), Mumbai, INDIA.
Instantaneous Frequencies”, IEEE Transactions on Audio, Speech, and He ha guided 17 Ph.D.s, 150 M.E./M.Tech Projects and several B.E./B.Tech
Language Processing, vol., 16, no. 6, August 2008. Projects. His areas of interest are Digital Signal processing, Image Processing
[10] Zhong-Xuan, Yuan & Bo-Ling, Xu & Chong-Zhi, Yu. (1999). “Binary and Computer Networks. He has more than 300 papers in National /
Quantization of Feature Vectors for Robust Text-Independent Speaker International Conferences / Journals to his credit. Recently twelve students
Identification” in IEEE Transactions. working under his guidance have received best paper awards. Recently two
[11] Dr. H B Kekre, Vaishali Kulkarni,”Speaker Identification using Power research scholars have received Ph. D. degree from NMIMS University
Distribution in Frequency Spectrum”, Technopath, Journal of Science, Currently he is guiding ten Ph.D. students. He is member of ISTE and IETE.
Engineering & Technology Management, Vol. 02, No.1, January 2010.
[12] Dr. H B Kekre, Vaishali Kulkarni, “Speaker Identification by using Vaishali Kulkarni has received B.E in Electronics
Power Distribution in Frequency Spectrum”, ThinkQuest - 2010 Engg. from Mumbai University in 1997, M.E (Electronics
International Conference on Contours of Computing Technology”, and Telecom) from Mumbai University in 2006. Presently
BGIT, Mumbai,13th -14th March 2010. she is pursuing Ph. D from NMIMS University. She has a
[13] H B Kekre, Vaishali Kulkarni, “Speaker Identification by using Vector teaching experience of more than 8 years. She is Associate
Quantization”, International Journal of Engineering Science and Professor in telecom Department in MPSTME, NMIMS
Technology, May 2010. University. Her areas of interest include Speech
processing: Speech and Speaker Recognition. She has 10 papers in National /
[14] H B Kekre, Vaishali Kulkarni, “Performance Comparison of Speaker
International Conferences / Journals to her credit.
Recognition using Vector Quantization by LBG and KFCG ” ,
International Journal of Computer Applications, vol. 3, July 2010.
[15] H B Kekre, Vaishali Kulkarni, “ Performance Comparison of
Automatic Speaker Recognition using Vector Quantization by LBG
KFCG and KMCG”, International Journal of Computer Science and
Security, Vol: 4 Issue: 5, 2010.
[16] H B Kekre, Vaishali Kulkarni, “Comparative Analysis of Automatic
Speaker Recognition using Kekre’s Fast Codebook Generation
Algorithm in Time Domain and Transform Domain ” , International
Journal of Computer Applications, Volume 7 No.1. September 2010.

143 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 3, March 2011

Reliability and Security in MDRTS
A Combine Colossal Expression

Gyanendra Kumar Gupta A. K Sharma Vishnu Swaroop
Computer Sc. & Engg. Deptt. Computer Sc. & Engg. Deptt Computer Sc. & Engg. Deptt
Kanpur Institute of Technology M.M.M. Engineering College M.M.M. Engineering College
Kanpurr, UP, India, 208001 Gorakhpur, UP, India, 273010 Gorakhpur, UP, India, 273010
gyanendrag@gmail.com akscse@rediffmail.com rsvsgkp@rediffmail.com

Abstract—Numerous types of Information Systems are broadly
used in various fields. With the fast development of computer I. INTRODUCTION (HEADING 1)
network, Information System users care more about data Data reliability summarizes the validity, accuracy,
sharing in networks. Sharing of information and changes made usability and integrity of related data between applications
by dissimilar user at different permission level is controlled by and across Information Technology. This ensures that each
super user, but the read/write operation is performed in a
user observes a reliable view of the constant data, including
reliable manner. In conventional relational database, data
reliability was controlled by consistency control mechanism
visible changes made by the user's own transactions
when a data object is locked in a sharing mode, other (read/write) and transactions of other users or processes
transactions can only read it, but can not update it. If the [1,2]. Data Reliability problems may arise at any time but are
conventional consistency control method has been used yet, the frequently introduced during or following recovery situations
system’s concurrency will be inadequately influenced. So there when backup copies of the data are used in place of the
are many new necessities for the consistency control in the field original data. Reliability is mostly concerned with
of Information system (MDRTS). In present era not only the consistency [3].
information grows enormously it also brings together in
Building distributed database system reliability is very
different nature of data like text, image, and picture, graphic
and sound. The problem not limited only to type of data (e.g.
important. The failure of a distributed database system can
databases) it has used in different environment of database like result in anything from easily repairable errors to disastrous
Mobile Database, Distributed, Real Time Database, and meltdowns. A reliable distributed database system is
Database and Multimedia database. There are many aspects of designed to be as fault tolerant as feasible. Fault tolerance
data reliability problems in mobile distributed real time system deals with making the system function in the presence of
(MDRTS), such as inconsistency between attribute and type of faults. Faults can occur in any of the components of a
data; the inconsistency of topological relations after objects has distributed system. This article gives a brief overview of the
been modified. In this paper, many cases of data reliability are different types of faults in a system and some of their
discussed for Information System. As the mobile computing solutions.
becomes well-liked and the database grows with information
sharing security is a big issue for researchers. Reliability and Various kinds of data consistency have been identified.
Security of data is a big confront for researchers because when These include Application Consistency, Transaction
ever the data is not reliable and secure no maneuver on the Consistency and Point-in-Time Consistency
data (e.g. transaction) is useful. It becomes more and more
crucial when the data changes from one form to another (i.e. II. VARIUOS TYPE OF CONSISTENCY
transactions) that are used in non-traditional environment like
Mobile, Distributed, Real Time and Multimedia databases. In A. Point in Time Consistency
this paper we raise the different aspects and analyze the
available solution for reliability and security of databases. Data is said to be Point in Time consistent if all of the
Conventional Database Security has focused primarily on interrelated data components are as they were at any single
creating user accounts and managing user privileges level to instant in time. This type of consistency can be visualized by
database objects. In this paper we also talk about an picturing a data center that has experienced a power failure.
impression of the present and past database security Before the lights come back on and processing resumes, the
challenges. data is considered time consistent, due to the fact that the
entire processing environment failed at the same instant of
Key Words- System Reliability, Sharing , Data time.
Consistency, Data Privileges, Data Loss, Data Recovery, Different types of failures may create a situation where
Integrity, Concurrency Control & Recovery, Distributed Point in Time consistency is not maintained. For example,
Databases, Transactions, Security, Authentication, consider the failure of a single logical volume containing
Integrity, Access Control, Encryption data from several applications. If the only recovery option is
to restore that volume from a backup taken sometime earlier,
the data contained on the restored volume is not consistent

144 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 3, March 2011
with the other volumes, and additional recovery steps must By and large we take synchronization features for granted
be undertaken.[101] and do not give much thought to how they all work together
to protect both the integrity and consistency of the data. It is
B. Transaction Consistency the integrity of the data and various systems that allows
A transaction is a logical unit of work that may include applications to restart after a power failure or other
any number of file or database updates. During normal unscheduled event.
processing, transaction consistency is present only
III. DATA LOSS VS. DATA CONSISTENCY
• Before any transactions have run. How does one reconcile the possibility of lost data versus
the integrity and consistency of the data? Often times,
• Follow the completion of a successful transaction traditional backups were created while files were being
and before the next transaction begins, and updated. Eventually, backups created in this fashion were
referred to as “fuzzy backups” as neither the consistency nor
• When the application ends normally or the the integrity of the data could be assured.
database is closed.
Importantly it is better idea to capture as many updates as
A failure of some kind, the data will not be transaction possible, even if the end result is not consistent. Let us
consistent if transactions were in-flight at the time of the consider this point within the confines of a "typical" large
failure. In most cases what occurs is that once the application systems data center. For the sake of discussion, let us assume
or database is restarted, the incomplete transactions are that there are many applications sharing data on hundreds of
identified and the updates relating to these transactions are logical volumes in many thousands of data sets. What
either “backed-out” or processing resumes with the next happens to the integrity of the data if some updates are
dependant write [4]. applied and others are not? Should this occur, the data is in
an artificial state, one that is neither time, transaction nor
C. Application Consistency application consistent? When the applications are restarted, it
It is similar to Transaction consistency, but on a grander is likely that some data will be duplicated, while other data
scale. Instead of data consistency within the scope of a single will still be missing. The difficulty here is in identifying
transaction, data must be consistent within the confines of which updates were successful, which updates caused
many different transaction streams from one or more erroneous results and which updates are missing.
applications. An application may be made up of many In all cases it is preferable to have time consistent data,
different types of data, such as multiple database even if a few partial transactions are lost or rolled back in the
components, various types of files, and data feeds from other process.
applications. Application consistency is the state in which all
related files and databases are in-synch and represent the true Data loss can be defined as data that is lost and cannot be
status of the application. recovered by another means. Often, individual transactions
or files can be restored or recreated, which is inconvenient,
Data Consistency refers to the usability of data and is but does not represent a true loss of data. Even in cases
often taken for granted in the single site environment. Data where some transactional data cannot be recreated or
Consistency problems may arise even in a single-site recovered by the data center support teams, it can sometimes
environment during recovery situations when backup copies be re-entered by the end user if necessary.
of the production data are used in place of the original data
[5]. If considering an asynchronous Business Continuity and
Disaster Recovery solution, it is important to understand that
In order to ensure that your backup data is useable, it is some updates may be lost in flight. However, the greater
necessary to understand the backup methodologies that are in consideration is that the asynchronous solution you select
place as well as how the primary data is created and provides you time consistent data for all of your interrelated
accessed. Another very important consideration is the applications. In this way, recovery is similar to the process
consistency of the data once the recovery has been completed necessary to achieve Transaction and Application
and the application is ready to begin processing. Consistency following an outage at the primary site.
In order to appreciate the integrity of your data, it is Data loss does not imply a loss of data integrity.
important to understand the dependent write process. This However, given a choice, most organizations will protect
occurs within individual programs, applications, application data consistency—for example, ensuring that bank deposits
systems and databases. A dependent write is a data update and withdrawals occur in the proper sequence so that account
that is not to be written until a previous update has been balances reflect a consistent picture any given point in time.
successfully completed. In the large systems environments, This is preferable to processing transactions out of sequence,
the logic that determines the sequence in which systems issue or, to use our banking example again, to record the
“writes” is controlled by the application processing flow and withdrawal and not the preceding deposit [7].
supported by basic system functions [6].

145 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 3, March 2011
IV. THE BACKUP PROBLEM - AN OVERVIEW It is true that different records would have been backed
For a set of backup data to be of any value it needs to be up if the write I/O pattern had been different, or if the backup
consistent in some fashion; Time, Transaction or Application process was either faster or slower. The point here is that
consistency is required. For an individual data set, one with unless the backup could have been processed instantaneously
no dependencies on any other data, this can be accomplished (or at least in the time between two of the file write I/Os), the
by creating a simple Point in Time copy of the data and backup copy does not represent consistent data within the
ensuring that the data is not updated during the backup file.
process.[8] In order to address this failing, various methods were
At peek, this appears to be a relatively simple thing to developed including transaction logging, transaction back out
accomplish -- at least for an individual data set. However, if and file reload with applied journal transactions, just to name
this data set is being updated by a critical on-line application, a few. These methods are all share the attributes of requiring
there may never be an opportunity to create a consistent extra effort (before the backup) and additional time -
backup-copy without temporarily halting the critical possibly even manual intervention - before the data can be
application. With today's dependence on 24x7 processing, used. More importantly, the corrective process requires an
the opportunities for even temporarily interrupting critical in-depth understanding of both the application and data.
applications to create a window” are seldom available [9]. These requirements dictate that a unique recovery scenario
be designed for nearly each and every data set.
As this problem became more prevalent, there were
various methods used to attempt to address the situation. One The integrity problem is daunting enough when viewed
of these methods was to create a “fuzzy” backup of the data, in the context of just these 20 records, but what about when
that is, to create the backup copy while updates were allowed there are interdependencies between thousands of data sets
to continue. Various utilities were used to perform this residing on hundreds (or even thousands) of volumes?
“backup while open” (BWO), but they all shared the attribute In this greater context, simple data consistency within
that the backup copy of the data may or may not be useable: individual data sets is no longer sufficient. What is required
If no additional actions were taken to validate and ensure the is time consistency across all of the interdependent data. As
consistency of the data, any use of this backup data was it is impossible to achieve this with the traditional backup
predicated on the hope that “some data is better than methodologies, newer technologies are required to support
nothing” and generally produced unpredictable and/or un- time consistent data?
repeatable results.
Fortunately, there are solutions available today. For a
In fact, there are three different possible outcomes, should single-site solution, FlashCopy with Consistency Groups can
this fuzzy backup be restored: be used to create a consistent Point-in-Time copy that can
then be backed-up by traditional means [11].
1. The data is accidentally consistent and useable. To guarantee the correct results and consistency of
This is a happy circumstance that may or may not databases, the conflicts between transactions can be either
be repeatable. avoided, or detected and then resolved. Most of the existing
mobile database CC techniques use the (conflict)
2. The data is not consistent and not useable. A serializability as the correctness criterion. They are either
subsequent attempt to use the data detects the pessimistic if they avoid conflicts at the beginning of
errors and abnormal end subsequent processing. transactions, or optimistic if they detect and resolve conflicts
right before the commit time, or hybrid if they are mixed. To
3. The data is NOT consistent, but does not cause an fulfill this goal, locking, timestamp ordering (TO) and
ABEND and happens to be useable to the serialization graph testing can be used as either a pessimistic
application. It is used by subsequent processing and or optimistic algorithm.
any data errors go undetected and uncorrected. This
is the worst possible outcome. V. SECURITY IN DATABASES
One of the first things it might be notice when looking at Database security is the system, processes, and
the records contained on the backup is that they are different procedures that protect a database from unintended activity.
from the data records that were present on the file both Unintended activity can be categorized as authenticated
before the backup started and immediately after the backup misuse, malicious attacks or inadvertent mistakes made by
ended. In fact, the records contained within the backup are a authorized individuals or processes. Database security is also
completely artificial construct and does not accurately a specialty within the broader discipline of computer
describe the contents of the file at any Point in Time. This is security. Databases introduce a number of unique security
not a consistent backup of the data. It is neither data- requirements for their users and administrators. On one hand,
consistent within itself nor is it time-consistent from any databases are designed to promote open and flexible access
point in time. It is a completely artificial representation of a to data. On the other hand, it’s this same open access that
file that never existed. [10] makes databases vulnerable to many kinds of malicious
activity. These are just a few of the database security
problems that exist within organizations. The best way to

146 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 3, March 2011
avoid a lot of these problems is to employ qualified your physical features, your voice, and fingerprint locks can
personnel and separate the security responsibilities from the read your fingerprints [14, 15].
daily database maintenance responsibilities [12, 31].
Access control is a rapidly growing market and soon may
Traditionally databases have been protected from manifest itself in such ways we cannot even imagine.
external connections by firewalls or routers on the network Nowadays, security access control is a necessary component
perimeter with the database environment existing on the for businesses. There are many ways to create this security.
internal network opposed to being located within a Some companies hire a security guard to stand in the
demilitarized zone. Additional network security devices that gateway. There are many security devices that prevent or
detect and alert on malicious database protocol traffic permit access such as a turnstile. The best most effective
include network intrusion detection systems along with host- access control systems are operated by computers.
based intrusion detection systems.
Auditing is a computer security audit is a manual or
One of the main issues faced by database security systematic measurable technical assessment of a system or
professionals is avoiding inference capabilities. Basically, application. Manual assessments include interviewing staff,
inference occurs when users are able to piece together performing security vulnerability scans, reviewing
information at one security level to determine a fact that application and operating system access controls, and
should be protected at a higher security level. Database analyzing physical access to the systems. Automated
security is more critical as networks have become more assessments include system generated audit reports or using
open. software to monitor and report changes to files and settings
on a system. Systems can include personal computers,
Databases provide many layers and types of information servers, mainframes, network routers, switches. Applications
security, typically specified in the data dictionary, including: can include Web Services, Databases [16].
Authentication is the process of confirming a user or
• Access control computer’s identity. The process normally consists of four
steps:
• Auditing

• Authentication 1. The user makes a claim of identity, usually by providing a
username. For example, It might make this claim by telling a
• Encryption database that my username is something.
2. The system challenges the user to prove his or her
• Integrity controls identity. The most common challenge is a request for a
Database security can begin with the process of creation password.
and publishing of appropriate security standards for the
3. The user responds to the challenge by providing the
database environment. The standards may include specific
requested proof. In this example, It would provide the
controls for the various relevant database platforms; a set of
database with my password.
best practices that cross over the platforms; and linkages of
the standards to higher level polices and governmental 4. The system verifies that the user has provided
regulations. acceptable proof by, for example, checking the password
against a local password database or using a centralized
Access Control is a term taken from the linguistic world
authentication server
of security. In general, it means the execution of limitations
and constrictions on whoever tries to occupy a certain Encryption is good. It helps make things more secure.
protected property. Guarding an entrance of a person is also a However, the idea that strong cryptography is good security
practice of access control. There are many types of access by itself is simply wrong. Encrypted messages eventually
control.[28]. Some of them are mentioned in this article. have to be decrypted so they are useful to the sender or
You, the reader of this article, will have several types of receiver. If those end-points are not secured, then getting the
access control around you. Nowadays, almost every plain-text messages is trivial [17]. This is a demonstration of
computer user has a firewall or antivirus is running on every a crude process of accomplishing that. There is no dispute
computer, a popup blocker and many other programs. All of about the need for strong encryption, particularly for
these are with access control functions [13]. All of these privileged communications. There is no way to have a high
programs guard us from intruders of sorts. They inspect level of assurance that the entire path between endpoints of a
everything trying to enter the computer and let it in or leave message is secure, so the message has to be hidden in transit.
it out. Computers have complicated access control abilities. While brute-force decryption is possible, modern forms of
They ask for authentication and search for the digital encryption have made this process too long to be valuable
signatures. Also, there are different types of keypads and [18].
access control systems. In today's world the keys and locks
are beginning to look different. With the passage of time, the Computer security authentication means verifying the
key locks also got smarter. They can identify the patterns of identity of a user logging onto a network. Passwords, digital
certificates, smart cards and biometrics can be used to prove

147 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 2, March 2011
the identity of the user to the network. Computer security database security manager also provides different types
authentication includes verifying message integrity, e-mail of access control for different users and assesses new
authentication and MAC (Message Authentication Code), programs that are performing with the database. If these
checking the integrity of a transmitted message. There are tasks are performed on a daily basis, you can avoid a lot
human authentication, challenge-response authentication, of problems with users that may pose a threat to the
password, digital signature, IP spoofing and biometrics [19, security of the database.
26].
Human authentication is the verification that a person • Varied Security Methods for Applications: More often
initiated the transaction, not the computer. Challenge- than not applications developers will vary the methods
response authentication is an authentication method used to of security for different applications that are being
prove the identity of a user logging onto the network. When utilized within the database. This can create difficulty
a user logs on, the network access server (NAS), wireless with creating policies for accessing the applications.
access point or authentication server creates a challenge, The database must also possess the proper access
typically a random number sent to the client machine. The controls for regulating the varying methods of security
client software uses its password to encrypt the challenge otherwise sensitive data is at risk.
through an encryption algorithm or a one-way hash function
and sends the result back to the network. This is the • Post-Upgrade Evaluation: When a database is upgraded
response. it is necessary for the administrator to perform a post-
Two- factor authentication requires two independent upgrade evaluation to ensure that security is consistent
ways to establish identity and privileges. The method of across all programs. Failure to perform this operation
using more than one factor of authentication is also called opens up the database to attack.
strong authentication. This contrasts with traditional
password authentication, requiring only one factor in order to • Split the Position: Sometimes organizations fail to split
gain access to a system. Password is a secret word or code the duties between the IT administrator and the
used to serve as a security measure against unauthorized database security manager. Instead the company tries to
access to data. It is normally managed by the operating cut costs by having the IT administrator do everything.
system or DBMS. However, a computer can only verify the This action can significantly compromise the security of
legality of the password, not the legality of the user. the data due to the responsibilities involved with both
The two major applications of digital signatures are for positions. The IT administrator should manage the
setting up a secure connection to a website and verifying the database while the security manager performs all of the
integrity of files transmitted. IP spoofing refers to inserting daily security processes.
the IP address of an authorized user into the transmission of
an unauthorized user in order to gain illegal access to a • Application Spoofing: Hackers are capable of creating
computer system. applications that resemble the existing applications
connected to the database. These unauthorized
Biometrics is a more secure form of authentication than applications are often difficult to identify and allow
typing passwords or even using smart cards that can be hackers access to the database via the application in
stolen. However, some ways have relatively high failure
disguise.
rates. For example, fingerprints can be captured from a water
glass and fool scanners.
• Manage User Passwords: Sometimes IT database
security managers will forget to remove IDs and access
VI. DATABASE SECURITY ISSUES: DATABASE SECURITY privileges of former users which leads to password
PROBLEMS AND HOW TO AVOID THEM vulnerabilities in the database. Password rules and
A database security manager is the most important asset maintenance needs to be strictly enforced to avoid
to maintaining and securing sensitive data within an opening up the database to unauthorized users.
organization. Database security managers are required to
multitask and juggle a variety of headaches that accompany • Windows OS Flaws: Windows operating systems are
the maintenance of a secure database. For any organization it not effective when it comes to database security. Often
is important to understand some of the database security theft of passwords is prevalent as well as denial of
problems that occur within an organization and how to avoid service issues. The database security manager can take
them. If it is understand that how, where, and why of precautions through routine daily maintenance checks.
database security you can prevent future problems from
occurring [20]. As organizations increase their reliance on, possibly
distributed, information systems for daily business, they
• Regular Maintenance: Database audit logs require daily become more vulnerable to security breaches even as they
review to make certain that there has been no data gain productivity and efficiency advantages. Though a
misuse. This requires overseeing database privileges number of techniques, such as encryption and electronic
signatures, are currently available to protect data when
and then consistently updating user access accounts. A
transmitted across sites, a truly comprehensive approach for

148 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 3, March 2011
data protection must also include mechanisms for enforcing transaction has inserted additional rows that satisfy
access control policies based on data contents, subject the condition.
qualifications and characteristics, and other relevant
contextual information, such as time. It is well understood VIII. INTRODUCTION TO DATA CONCURRENCY AND
today that the semantics of data must be taken into account in CONSISTENCY IN A MULTIUSER ENVIRONMENT
order to specify effective access control policies. Also,
techniques for data integrity and availability specifically In a single-user database, the user can modify data in the
tailored to database systems must be adopted. In this respect, database without concern for other users modifying the same
over the years the database security community has data at the same time. However, in a multiuser database, the
developed a number of different techniques and approaches statements within multiple simultaneous transactions can
to assure data confidentiality, integrity, and availability. update the same data. Transactions executing at the same
However, despite such advances, the database security area time need to produce meaningful and consistent results.
faces several new challenges. Factors such as the evolution Therefore, control of data concurrency and data consistency
of security concerns, the "disintermediation¿ of access to is vital in a multiuser database [22].
data, a new computing paradigms and applications, such as
grid-based computing and on-demand business. we have • Data concurrency means that many users can access
introduced both new security requirements and new contexts data at the same time.
in which to apply and possibly extend current approaches. In
this review, we first survey the most relevant concepts • Data consistency means that each user sees a
underlying the notion of database security and summarize the consistent view of the data, including visible
most well-known techniques. We focus on access control changes made by the user's own transactions and
systems, on which a large body of research has been devoted, transactions of other users.
and describe the key access control models, namely, the
discretionary and mandatory access control models, and the To describe consistent transaction behavior when
role-based access control model. We also discuss security for transactions execute at the same time, database researchers
advanced data management systems, and cover topics such have defined a transaction isolation model called
as access control for XML. We then discuss current serializability. The serializable mode of transaction behavior
challenges for database security and some preliminary tries to ensure that transactions execute in such a way that
approaches that address some of these challenges [21]. they appear to be executed one at a time, or serially, rather
than concurrently [31].
VII. MAJOR SECURITIES CHALLENGES While this degree of isolation between transactions is
generally desirable, running many applications in this mode
can seriously compromise application throughput. Complete
1. Security Awareness and End-users
isolation of concurrently running transactions could mean
2. Google Exposure
that one transaction cannot perform an insert into a table
3. Standard Compliance & Regulations Updates being queried by another transaction. In short, real-world
4. Vulnerability Management considerations usually require a compromise between perfect
5. Frequently Change of Management and Lack of Co- transaction isolation and performance.
ordination in Management
In general, multiuser databases use some form of data
Review the four levels of transaction isolation with locking to solve the problems associated with data
differing degrees of impact on transaction processing concurrency, consistency, and integrity. Locks are
throughput. These isolation levels are defined in terms of mechanisms that prevent destructive interaction between
three phenomena that must be prevented between transactions accessing the same resource.
concurrently executing transactions.
Resources include two general types of objects:
The three preventable phenomena are:
• User objects, such as tables and rows (structures
• Dirty reads: A transaction reads data that has been and data)
written by another transaction that has not been
committed yet. • System objects not visible to users, such as shared
data structures in the memory and data dictionary
• Non-repeatable (fuzzy) reads: A transaction rereads rows
data it has previously read and finds that another
committed transaction has modified or deleted the Database automatically provides read consistency to a
data. query so that all the data that the query sees comes from a
single point in time (statement-level read consistency).
• Phantom reads: A transaction re-executes a query Database can also provide read consistency to all of the
returning a set of rows that satisfies a search queries in a transaction (transaction-level read consistency).
condition and finds that another committed

149 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 32, March 2011
Database uses the information maintained in its rollback this situation by setting higher values of INITRANS for
segments to provide these consistent views. The rollback tables that will experience many transactions updating the
segments contain the old values of data that have been same blocks. Doing so enables Database to allocate sufficient
changed by uncommitted or recently committed transactions. storage in each block to record the history of recent
Database provides statement-level read consistency using transactions that accessed the block.
data in rollback segments.
Database generates an error when a serializable
1) Statement-Level Read Consistency transaction tries to update or delete data modified by a
Database always enforces statement-level read transaction that commits after the serializable transaction
consistency. This guarantees that all the data returned by a began.When a serializable transaction fails with the "Cannot
single query comes from a single point in time--the time that serialize access" error, the application can take any of several
the query began. Therefore, a query never sees dirty data nor actions:
any of the changes made by transactions that commit during
query execution. As query execution proceeds, only data • Commit the work executed to that point
committed before the query began is visible to the query. The
query does not see changes committed after statement • Execute additional (but different) statements
execution begins. (perhaps after rolling back to a save point
2) Read Consistency with Real Application Clusters established earlier in the transaction)
Real Application Clusters use a cache-to-cache block
transfer mechanism known as Cache Fusion to transfer read- • Roll back the entire transaction
consistent images of blocks from one instance to another. 5) Comparison of Read Committed and Serializable
Real Application Clusters does this using high speed, low
Isolation
latency interconnects to satisfy remote requests for data
Database gives the application developer a choice of two
blocks.
transaction isolation levels with different characteristics.
3) Read Committed Isolation Both the read committed and serializable isolation levels
The default isolation level for Database is read provide a high degree of consistency and concurrency. Both
committed. This degree of isolation is appropriate for levels provide the contention-reducing benefits of Database's
environments where few transactions are likely to conflict. read consistency multiversion concurrency control model
Database causes each query to execute with respect to its and exclusive row-level locking implementation and are
own materialized view time, thereby permitting designed for real-world application deployment.
nonrepeatable reads and phantoms for multiple executions of
a query, but providing higher potential throughput. Read a) Transaction Set Consistency
committed isolation is the appropriate level of isolation for A useful way to view the read committed and serializable
environments where few transactions are likely to conflict. isolation levels in Database is to consider the following
scenario: Assume you have a collection of database tables (or
4) Serializable Isolation any set of data), a particular sequence of reads of rows in
Serializable isolation is suitable for environments: those tables, and the set of transactions committed at any
particular time. An operation (a query or a transaction) is
transaction set consistent if all its reads return data written by
• With large databases and short transactions that
the same set of committed transactions. An operation is not
update only a few rows
transaction set consistent if some reads reflect the changes of
one set of transactions and other reads reflect changes made
• Where the chance that two concurrent transactions by other transactions. An operation that is not transaction set
will modify the same rows is relatively low consistent in effect sees the database in a state that reflects no
single set of committed transactions.
• Where relatively long-running transactions are
primarily read-only Database provides transactions executing in read
committed mode with transaction set consistency for each
Serializable isolation permits concurrent transactions to statement. Serializable mode provides transaction set
make only those database changes they could have made if consistency for each transaction.
the transactions had been scheduled to execute one after
another. Specifically, Database permits a serializable b) Row-Level Locking
transaction to modify a data row only if it can determine that Both read committed and serializable transactions use
prior changes to the row were made by transactions that had row-level locking, and both will wait if they try to change a
committed when the serializable transaction began. row updated by an uncommitted concurrent transaction. The
Under some circumstances, Database can have second transaction that tries to update a given row waits for
insufficient history information to determine whether a row the other transaction to commit or roll back and release its
has been updated by a "too recent" transaction. This can lock. If that other transaction rolls back, the waiting
occur when many transactions concurrently modify the same transaction, regardless of its isolation mode, can proceed to
data block, or do so in a very short period. You can avoid

150 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 3, March 2011
change the previously locked row as if the other transaction and serializable isolation provide a high level of concurrency
had not existed. for high performance, without the need for reading
uncommitted ("dirty") data [23, 24]
However, if the other blocking transaction commits and
releases its locks, a read committed transaction proceeds with e) Read Committed Isolation
its intended update. A serializable transaction, however, fails For many applications, read committed is the most
with the error "Cannot serialize access", because the other appropriate isolation level. Read committed isolation can
transaction has committed a change that was made since the provide considerably more concurrency with a somewhat
serializable transaction began. increased risk of inconsistent results due to phantoms and
c) Referential Integrity non-repeatable reads for some transactions.
Because Database does not use read locks in either read- Many high-performance environments with high
consistent or serializable transactions, data read by one transaction arrival rates require more throughput and faster
transaction can be overwritten by another. Transactions that response times than can be achieved with serializable
perform database consistency checks at the application level isolation. Other environments that support users with a very
cannot assume that the data they read will remain unchanged low transaction arrival rate also face very low risk of
during the execution of the transaction even though such incorrect results due to phantoms and no repeatable reads.
changes are not visible to the transaction. Database Read committed isolation is suitable for both of these
inconsistencies can result unless such application-level environments.
consistency checks are coded with this in mind, even when
using serializable transactions. Database read committed isolation provides transaction
set consistency for every query. That is, every query sees
d) Distributed Transactions data in a consistent state. Therefore, read committed isolation
In a distributed database environment, a given transaction will suffice for many applications that might require a higher
updates data in multiple physical databases protected by two- degree of isolation if run on other database management
phase commit to ensure all nodes or none commit. In such an systems that do not use multiversion concurrency control.
environment, all servers, whether Database or non-Database, Read committed isolation mode does not require
that participates in a serializable transaction are required to application logic to trap the "Cannot serialize access" error
support serializable isolation mode. and loop back to restart a transaction. In most applications,
few transactions have a functional need to issue the same
If a serializable transaction tries to update data in a database query twice, so for many applications protection against
managed by a server that does not support serializable phantoms and non-repeatable reads is not important.
transactions, the transaction receives an error. The Therefore many developers choose read committed to avoid
transaction can roll back and retry only when the remote the need to write such error checking and retry code in each
server does support serializable transactions. transaction.
In contrast, read committed transactions can perform f) Serializable Isolation
distributed transactions with servers that do not support Database's serializable isolation is suitable for
serializable transactions. environments where there is a relatively low chance that two
Application designers and developers should choose an concurrent transactions will modify the same rows and the
isolation level based on application performance and long-running transactions are primarily read-only. It is most
consistency needs as well as application coding suitable for environments with large databases and short
requirements. transactions that update only a few rows.
For environments with many concurrent users rapidly Serializable isolation mode provides somewhat more
submitting transactions, designers must assess transaction consistency by protecting against phantoms and
performance requirements in terms of the expected nonrepeatable reads and can be important where a read/write
transaction arrival rate and response time demands. transaction executes a query more than once.
Frequently, for high-performance environments, the choice Unlike other implementations of serializable isolation,
of isolation levels involves a trade-off between consistency which lock blocks for read as well as write, Database
and concurrency. provides nonblocking queries and the fine granularity of
Application logic that checks database consistency must row-level locking, both of which reduce write/write
take into account the fact that reads do not block writes in contention. For applications that experience mostly
either mode. read/write contention, Database serializable isolation can
provide significantly more throughput than other systems.
Database isolation modes provide high levels of Therefore, some applications might be suitable for
consistency, concurrency, and performance through the serializable isolation on Database but not on other systems.
combination of row-level locking and Database's
multiversion concurrency control system. Readers and Coding serializable transactions requires extra work by
writers do not block one another in Database. Therefore, the application developer to check for the "Cannot serialize
while queries still see consistent data, both read committed access" error and to roll back and retry the transaction.

151 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. . 9, No. 3, March 2011
Similar extra coding is needed in other database management impair the integrity and availability of a database. [26, 27,
systems to manage deadlocks. For adherence to corporate 28]
standards or for applications that are run on multiple
database management systems, it may be necessary to design There are several technique have been built for
transactions for serializable mode. Transactions that check maintaining the security and reliability of systems like Data
for serializability failures and retry can be used with Consistency Techniques - Two-Process Mutual Exclusion:
Database read committed mode, which does not generate Dekker's- and Peterson's Algorithms, N-Process Mutual
serializability errors. Exclusion using Hardware, N-Reader, 1-Writer Mutual
Exclusion using Head/Tail Flags. But the available
Serializable mode is probably not the best choice in an techniques are not sufficient for the different database
environment with relatively long transactions that must environment where the data is huge and complex for
update the same rows accessed by a high volume of short transactions including security system. The unusual
update transactions. Because a longer running transaction is requirement in security, how ever mean that designers must
unlikely to be the first to modify a given row, it will careful consider their opinions when choosing database
repeatedly need to roll back, wasting work. Note that a technology for deployment commercially available products
conventional read-locking, pessimistic implementation of can provide outstanding performance, reliability, scalability
serializable mode would not be suitable for this environment but unless they are expressly for embedded use, may
either, because long-running transactions--even read compromise overall security system. Security is more than
transactions--would block the progress of short update Just Good Crypto - The point here is not that encryption is
transactions and vice versa.) worthless. The point is that encryption by itself is not helpful
[29]. The endpoints need to be secure, passwords need to be
Developers should take into account the cost of rolling difficult to crack, and those who do have access to the
back and retrying transactions when using serializable mode. system need to be trustworthy. One might ask what is the
As with read-locking systems, where deadlocks occur point of being able to see plaintext versions of encrypted
frequently, use of serializable mode requires rolling back the communication if they already have root access? Getting
work done by terminated transactions and retrying them. In a additional passwords for other systems, obtaining
high contention environment, this activity can use significant information that passes through the system but is not stored
resources. on the system (text conversations, for instance), or bypassing
For the most part in transaction operations, a transaction system controls that might catch direct attempts at data.
that restarts after receivi