Implementation of Principle Component Analysis with Fuzzy Annotation for CAD Jewellery Images

							    International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November – December 2012                                    ISSN 2278-6856




             Implementation of Principle Component
             Analysis with Fuzzy Annotation for CAD
                         Jewellery Images
                                                       Pinderjeet Kaur
                               Guru Nanak Dev Engineering College, Ludhiana (Punjab), India - 141006,
                                                Punjab Technical University, India

Abstract: It is not easy to look for a specific image in a        document such as book and journal articles. For
large database but it can be an interesting topic for research.   structured information we find answer and information in
It is well known that we face any issues in searching for         already existing systems that are database management
images in large databases using various algorithms. In this
                                                                  system.
research work I have developed a Content Based Image
Retrieval (CBIR) System for jewellery images. As it is true
                                                                  In information Retrieval Systems user enters a query for
that a lot of work has already been done in this field (CBIR)     information that he needs. The query is represented in the
for SAR images and in medical field, but none has been done       form of question. The System looks for the requested
for CAD jewellery images. The Content Based Image                 information in the whole document and in available
Retrieval (CBIR) system which I have developed works on the       resources; this is known as acquisition of documents and
principle of Principal Component Analysis (PCA). Principal        objects. In system indexing is used for representation of
Component Analysis (PCA) is used for dimension reduction
                                                                  documents. System selects the document and other objects
so that the computation cost for the system of Content Based
Image Retrieval (CBIR) will be reduced.                           from various sources. In this way query is matched with
Keywords: Content Based Image Retrieval (CBIR),                   indexed documents.
Principal Component Analysis (PCA), Image Retrieval
(IR), Computer Aided Design (CAD)

1. INTRODUCTION
Information retrieval (IR) term means that it is relate
with the search of structured information which is fit for
relational database, unstructured information which is not
fit for relational database, search for the documents,
information of the document, metadata about             the
documents       and     searching relational     databases.
Information retrieval system is a traditional model.
Information retrieval provides the user with effective
access and interaction with information resources. In an                       Figure 1: Information System
information retrieval system user can enter a query and
then system look for the query into the system. Queries              1.1 Objective of Image Retrieval System
are formal statement and it is not possible to recognize a        Traditionally Information Retrieval System has been used
particular object by the system, many objects may match           to retrieve text information from various images but these
with the query. An object is an information that is need          days with the help of this Information system we can
by the user and which is matched with the collection of           retrieve information in many other interesting areas:-
data in the information retrieval system. Information              Speech retrieval which deals with speech
retrieval system is a part of a family that shares many             reorganization by automated speech reorganization.
principles. In an Information retrieval system search is           For image retrieval we can find image by giving its
related with unstructured information and structured                shape, color and size.
information. For unstructured information we find answer           For music retrieval we can find a musical file by giving
and information in already exiting two systems that is              a musical theme.
hypermedia system and information retrieval system.
Hypermedia system deals with many small units such as             2. WHAT IS CBIR?
paragraph and single images which are tied together by            Content based image retrieval (CBIR) is the retrieval of
links. Information retrieval system deals with the whole          images that matches the user query by analyzing the
Volume 1, Issue 4 November - December 2012                                                                         Page 87
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November – December 2012                                    ISSN 2278-6856


visual content of the images in the database. When a           descriptors to describe image content, as this can provide
query is made features like shape, colour and texture are      a greater degree of language independence and define
searched to find the exact match to the features queried by    concept hierarchies more vividly.[1] When discussing the
the user.                                                      indexing of images, we should distinguish between
There are two methods to retrieve an image from large          systems which work on formal description of the image
amount of data one is text based and other one is content      and ones based on image contents. Fuzzy Color Base
based. Text based means that an image can be retrieve by       Annotation Scheme, based on content based indexing,
textual information like file name, keywords, tags and         can be useful in building a more versatile Content Based
                                                               Image Retrieval system that can handle the usual
descriptions. We face many problems in using various
                                                               linguistic queries. My proposed work takes advantage of
methods to retrieve an image; image indexing is one of
                                                               Fuzzy Color Base Annotation Scheme. The query to
them, it is very difficult task to index images in a huge      retrieve the images from database is prepared in terms of
database of images. This needs a lot of human labor and        natural language such as mostly content, much content
it is time consuming process. This problem will help to        and few content of the some specific color.
develop a technique to retrieve an image from large
collection of images with the help of features such as            2.2 Principal Component Analysis (PCA)
colour, texture and shape. This technique to retrieve an       Principal component analysis (PCA) is a mathematical
image from large amount of data on the basis of                procedure or statistical technique for data contact and
automatically-derived features is referred as Content-         information extraction. [2] Principal Component Analysis
Based Image Retrieval (CBIR).                                  (PCA) is the general name for a technique which uses
Content-Based Image Retrieval (CBIR) means that the            sophisticated underlying mathematical principles to
search will focus on the actual contents of the image          transforms a number of possibly correlated variables into
rather than the tags or text related to it. [1] The term       a smaller number of variables called principal
‘content’ means colors, shapes, textures, or any other         components. Principal component analysis (PCA) is one
information that can be derived from the images. We            of the most important results from applied linear algebra.
have many web based search engines which depend on             It is used to reduce the dimension of datasets,
the metadata such as tags, keywords, file name and             classification, feature extraction, etc. Principal component
description. The result of these search engines is that they   analysis (PCA) uses a vector space transform to reduce
result in large amount of unwanted data, so we require         the dimensionality of large data sets.[3][4]Using
Content-Based Image Retrieval (CBIR). It is a better           mathematical projection, the original data set, which may
approach to retrieve an image with content based rather        have involved many variables, can often be interpreted in
than image indexing. Content-Based Image Retrieval             just a few variables (principal components).[3] Principal
(CBIR) infers many of its approaches from the field of         component analysis (PCA) is used to find the similar
image processing. Its approaches accent to retrieve an         images. Digital images are stored in the database. User’s
image with desired characteristics but its approaches          desire is to retrieve the similar image from the database.
differ from the field of image processing principally.         The process of annotation should be done properly to find
Image processing covers a vast area including Basic Gray       the desired and the similar image from the database.
Level Functions, Image enhancement, Image Averaging,           Fuzzy Color Base Annotation Scheme can be used for
Image Restoration, Image compression, Image                    image tagging and linguistic indexing. [4] In this way we
transmission, and Image interpretation. In image               can construct a metadata in form of caption or keywords
processing objects are recognized by feature analysis. The     for the database of images. Principal component analysis
difference between image analysis and CBIR is very much        (PCA) is a useful statistical technique which is commonly
clear. As we are using automatic face recognition              used for finding patterns in data of high dimension.
systems, this type of system may be used in two ways. The
first way is in which the image is compared with a single        2.3 Fuzzy Color Based Annotation Scheme
individual’s database record. The Second way is that, the      Fuzzy-based relevance feedback in Content Based Image
image is compared with the whole database to obtain the        Retrieval system shows the feasibility. Query image is
exactly matching images.                                       given as input images for retrieval for color based
                                                               retrieval. [5] My proposed work takes advantage of Fuzzy
  2.1 Classification of image retrieval system                 Color Base Annotation Scheme. The query to retrieve the
Current image retrieval techniques can be classified           images from database is prepared in terms of natural
according to the type and the nature of the features used      language such as mostly content, much content and few
for indexing. Probably the best-known indexing scheme          content of the some specific color. With the help of Fuzzy
in the public domain is the Art and Architecture               Color Base Annotation Scheme we can perform image
Thesaurus.[1] A good approach for indexing scheme is           tagging and linguistic indexing. In this way we can
using classification codes rather than keywords or subject

Volume 1, Issue 4 November - December 2012                                                                       Page 88
    International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November – December 2012                                    ISSN 2278-6856


construct a metadata in form of caption or keywords for           what is expected while executing human dialect and
the database of images. [6]                                       human linguistic descriptors based query, a system must
We define thirteen colors that fall within the human              be designed to be close to human dialect and linguistic
range perception. The feature representation set of colors        descriptors.
are (Black, Green, Yellow, Red, Cyan, Blue, White,
Neutral, Magenta, Flesh, Purple, Brown, and Pastel).
These thirteen colors are used as input. Color is most
                                                                  REFERENCES
popular feature which is used for image retrieval and for         [1] John Eakins Margaret Graham “Content-based
indexing. On the other hand, due to its basic nature of               Image Retrieval” JISC Technology Applications
error in description of the same semantic content by                  Programme. pp: 9, 16-17. (journal style)
different, color quantization and by the ambiguity of             [2] Jon Shlens “A Tutorial On Principal Component
human perception, it is important to capture this error               Analysis”     Discussion    and Singular       Value
when define the features. We apply fuzzy logic to the                 Decomposition pp:1-2 2003 (journal style)
traditional color histogram to help capture this ambiguity        [3] Hervé Abdi, Lynne J. Williams “Wiley
in image retrieval and in color indexing. [6]                         Interdisciplinary Reviews: Computational Statistics”
                                                                      Volume 2, Issue 4, pp 433–459, 2010 (book style)
3. CONCLUSION                                                     [4] Kim Esbensen, Paul Geladi “Chemometrics and
Fuzzy Color Base Annotation Scheme can be useful in                   Intelligent Laboratory Systems” Volume 2, Issues 1–
building a more reliable Content Based Image Retrieval                3, pp 37–52 August 1987 (book style)
system that can handle the usual linguistic queries. With         [5] Tamalika Chaira, A.K. Ray “Fuzzy Sets and
the help of Fuzzy Color Base Annotation Scheme we can                 Systems” Volume 150, Issue 3, pp. 545–560 2005
perform image tagging and linguistic indexing. In this                (book style)
way we can construct a metadata in form of caption or             [6] Jia Li “Real-Time Computerized Annotation of
keywords for the database of images.                                  Pictures” Pattren Analysis and Machinne Intelligence
                                                                      pp. 985-1002 2008 (journal style)
4. FUTURE WORK
                                                                  AUTHOR
As the CAD (Jewellery) images have very less contents
embedded in it, we still need to retrieve this information
                                                                                        Pinderjeet Kaur received Bachelor of
based on the user query. The query may include the image                                Technology (B.Tech) degree in Computer
itself and all possible descriptors of such images.                                     Science and Engineering (CSE) from
Therefore, any system to be built for such purposes must                                Institute of Engineering & Technology,
be close to human dialect and linguistic descriptors. Any                               Bhaddal (PTU) in 2009. She is pursuing
artistic, aesthetic, artifacts like jewellery need special case                         Masters of Technology (M.Tech) from
in terms of their semantics. Especially, if they are to be                              Guru Nanak Dev Engineering College
incorporate in information retrieval system which uses                                  Ludhiana (PTU). Pinderjeet Kaur has
                                                                  been working as a Lecturer since 2010 and is currently
content of images (pixel, numerical values etc) to built
                                                                  associated with Doaba Group of Collage, Kharar-Chandigarh.
and map with the human linguistic description of
jewellery item images and designs.
This leads to the creation of gap between the features of a
CAD (Jewellery) image being stored in a database and
what is expected while executing human dialect and
human linguistic descriptors based query e.g.
 Find jewellery images having red stones.
 Find jewellery images having round pendent with
   pearls.
Although the CAD (Jewellery) images have very less
contents embedded in it, we still need to retrieve this
information based on the user query.
However, for any technologist an image is nothing but a
group of pixels having certain mathematical features
which are hard for any human to comprehend easily until
he/she himself/herself know such technical buzzwords.

To reduce the semantic gap between the features of a
CAD (Jewellery) image being stored in a database and
Volume 1, Issue 4 November - December 2012                                                                          Page 89