High Security Human Recognition System using Iris Images
In this paper, efficient biometric security technique for Integer Wavelet Transform based Human Recognition System (IWTHRS) using Iris images verification is described. Human Recognition using Iris images is one of the most secure and authentic among the other biometrics. The Iris and Pupil boundaries of an Eye are identified by Integro-Differential Operator. The features of the normalized Iris are extracted using Integer Wavelet Transform and Discrete Wavelet Transform. The Hamming Distance is used for matching of two Iris feature vectors. It is observed that the values of FAR, FRR, EER and computation time required are improved in the case of Integer Wavelet Transform based Human Recognition System as compared to Discrete Wavelet Transform based Human Recognition System (DWTHRS).
ACEEE International Journal on Signal and Image Processing Vol 1, No. 1, Jan 2010 High Security Human Recognition System using Iris Images C. R. Prashanth1, Shashikumar D.R.2, K. B. Raja3, K. R. Venugopal3, L. M. Patnaik4 1 Department of Electronics and Communication Engineering, Vemana Institute of Technology, Bangalore, India 2 Department of Computer Science and Engineering, Cambridge Institute of Technology, Bangalore, India 3 Department of Computer Science and Engineering, University Visvesvaraya College of Engineering, Bangalore University, Bangalore 560 001, India 4 Vice Chancellor, Defence Institute of Advanced Technology, Pune, India email@example.com Abstract—In this paper, efficient biometric security after two or three years. (ii) The human Iris might be as technique for Integer Wavelet Transform based Human distinct as the Finger Prints for the different individuals. Recognition System (IWTHRS) using Iris images (iii) The forming of Iris depends on the initial verification is described. Human Recognition using Iris environment of the Embryo and hence the Iris Texture images is one of the most secure and authentic among the Pattern does not correlate with genetic determination. (iv) other biometrics. The Iris and Pupil boundaries of an Eye are identified by Integro-Differential Operator. The features Even the left and the right Irises of the same person are of the normalized Iris are extracted using Integer Wavelet unique. (v) It is almost impossible to modify the Iris Transform and Discrete Wavelet Transform. The Hamming structure by surgery. (vi) The Iris Recognition is non- Distance is used for matching of two Iris feature vectors. It invasive. (vii) It has about 245 degrees of freedom. is observed that the values of FAR, FRR, EER and Iris is the only internal organ which can be seen computation time required are improved in the case of outside the body. The probability of uniqueness among Integer Wavelet Transform based Human Recognition all humans has made Iris Recognition a reliable and System as compared to Discrete Wavelet Transform based efficient Human Recognition Technique. An Iris Human Recognition System (DWTHRS). biometric system can be utilized in two contexts: verification and identification. Verification is a one-to- Index Terms—Human Recognition, Biometrics, Integer Wavelet Transforms, Iris Image, High Security. one match in which the biometric system tries to verify a person’s identity by comparing the distance between test Iris and the corresponding Iris in the database, with a I. INTRODUCTION predefined threshold. If the computed distance is smaller Biometric solutions address the security issues than the predefined threshold, the subject is accepted as associated with traditional method of Human Recognition being genuine, else the subject is rejected. Identification based on personal identification number (PIN), identity is a one-to-many match in which the system compares the card, secrete password etc., and the traditional methods test Iris with all the Irises in the database and chooses the face severe problems such as loss of identity cards and sample with the minimum computed distance i e., forgetting/ guessing the passwords. Biometric measures greatest similarity as the identified result. If the test Iris based on physiological or behavioral characteristics are and the selected database Iris are from the same subject, it unique to an individual and have the ability to reliably is a correct match. The term authentication is often used distinguish between genuine person and an imposter. The as a synonym for verification. physiological characteristics include Iris, Finger Print, The Iris Verification system can be split into four Retinal, Palm Prints, Hand Geometry, Ear, Face and stages: data acquisition, segmentation, encoding and DNA, while the behavioral characteristics include matching. The data acquisition step captures the Iris Handwriting, Signature, Body Odor, Gait, Gesture and images using Infra-Red (IR) illumination. The Iris Thermal Emission of Human Body. Segmentation step localizes the Iris region in the image. The biometric systems based on behavioral For most algorithms and assuming near-frontal characteristics fail in many cases as the characteristics presentation of the Pupil, the Iris boundaries are modeled can easily be learnt and changed by practice. Some of the as two circles, which are not necessarily concentric. The techniques based on physiological characteristics such as inner circle is the pupillary boundary between the Pupil Face Recognition, Finger Prints and Hand Geometry also and the Iris whereas the outer circle is the limbic fail when used over a long time as they may change due boundary between the Iris and the Sclera. The noise due to ageing or cuts and burns. Among all the biometric to Eyelid occlusions, Eyelash occlusions, Specular techniques Iris Recognition has drawn a lot of interest in highlights and Shadows are eliminated using Pattern Recognition and Machine Learning research area segmentation. Most segmentation algorithms are gradient because of the advantages viz., (i) The Iris formation based that is segmentation is performed by finding the starts in the third month of gestation period and is largely Pupil-Iris edge and the Iris-Sclera edge. The encoding complete by the eighth month and then it does not change stage encodes the Iris image texture into a bit vector code. 26 © 2010 ACEEE DOI: 01.ijsip.01.01.06 ACEEE International Journal on Signal and Image Processing Vol 1, No. 1, Jan 2010 The corresponding matching stage calculates the distance Chin et al.,  proposed the use of an S-Iris encoding between Iris codes, and decides whether it is a match in which is generated from the inner product of the output the verification context or recognizes the submitted Iris from a 1D Log Gabor filter and secret pseudorandom from the subjects in the database. Biometrics is widely numbers. In the segmentation stage, first an edge map is used in many applications such as access control to generated using a Canny edge detector. A Circular Hough secure facilities, verification of financial transactions, Transform is used to obtain the Iris boundaries. Linear welfare fraud protection, law enforcement, and Hough Transform is used in excluding the Eyelid and immigration status checking when entering a country. Eyelash noises. The isolated Iris part is unwrapped into a Contribution: In this paper, we propose a novel rectangle with a resolution of 20 * 240 using Daugman’s technique for human identity authentication by Iris rubber sheet model. In matching, Hamming Distance is Verification. We use Integro-Differential equation for Iris used to indicate the dissimilarity between a pair of Iris localization and Daugman’s rubber-sheet model for codes. normalization. Integer Wavelet Transformation is used to Ya-Ping Huang et al.,  proposed a recognition extract the features from the normalized Iris image. method which constructs basic functions for training set Matching between the test image and the database images by Independent Component Analysis, which determines is done using Hamming Distance. the centre of each class by competitive learning Organization of the paper: The rest of the paper is mechanism and finally recognizes the pattern based on organized as follows. In section II, we discuss about Euclidean Distance. No restriction for image capture literature survey. In section III we present the Iris based owing to representation of size and rotation invariance. Human Recognition model. In section IV we discuss the However, the algorithm uses all patterns of each class as IWTHRS algorithm. The performance analysis presented a whole to estimate ICA basic function and when a new in section V and concluded in section VI. class is added all the patterns must be trained again. Schmid et al.,  proposed an algorithm to predict the II. LITERATURE SURVEY Iris Biometrics system performance on a larger dataset based on the Gaussian Model constructed from a smaller Daugman’s Algorithm [1, 2] proposed the Iris model dataset. In the matching stage, it uses a sequence of K Iris as two circles between the Pupil and Sclera boundaries, codes to represent an Iris subject. The distance between a which are not necessarily concentric. Each circle is pair of Iris subjects is defined as a K-dimensional defined by three parameters (xo, yo, r), where (xo, yo) Hamming Distance, modeled as Gaussian Distribution. locates the center of the circle of radius r. An Integro- Fancourt et al.,  discussed the problem of Iris Differential Operator is used to estimate the three Recognition using images acquired up to 10 meters away. parameter values for each circular boundary. The The pictures are captured with the aid of a telescope. The segmented Iris image is normalized and converted from manual Iris segmentation is used as a bootstrap to the Cartesian image coordinates to polar image coordinates. automatic segmentation. The similarity between the The 2D Gabor filter is used to encode the Iris image to a gallery image and probe image is measured by the binary code of 256 bytes in length. Hamming Distance is average correlation coefficient over sub-blocks with a used to verify the similarity of two Iris codes. size of 12*12 pixels. The algorithm is tested on two iris In an algorithm proposed by Ma et al., , the Iris databases with no subjects in common. images are projected to the vertical and horizontal directions to estimate the center of the Pupil, to save time III. IWTHRS MODEL in searching for the Iris boundaries. The region of Iris is constrained close to the Pupil, because Iris texture is In this section, IWTHRS model is discussed. Figure 1 claimed to be more abundant and also it reduces Eyelid shows the block diagram of Integer Wavelet Transform and Eyelash noise. The representation of the Iris is a based Human Recognition System (IWTHRS), which feature vector of length 1,536 bits. A Fisher Linear verifies the authenticity of given Iris of a person. The Eye Discriminant is used to reduce the dimension of the Iris images for study are taken from the CASIA database. The feature vector. Integro-Differential Operator (IDO) is used for Iris Kong and Zhang  proposed an Eyelash and localization and Daugman’s rubber-sheet model for reflection segmentation in their algorithm. The Iris normalization. Integer Wavelet Transformation is used to segmentation is implemented by using curve fitting extract the features from the normalized Iris image. approaches. The Eyelashes are sub-classified as separable Matching between the test Iris and the database Irises is Eyelashes and multiple Eyelashes. The separable done using Hamming Distance. Eyelashes are segmented using a Gabor filter and the A. Integro-Differential Operator for Image Segmentation multiple Eyelashes are segmented by comparing the variance of intensity values of a given area with the The Integro-Differential Operator is defined by the predefined threshold. Four types of 1-D wavelets viz., Equation 1. Mexican hat, Haar, Shannon and Gabor are used to ∂ I ( x, y ) max (r , xo , yo ) = Gσ (r ) ∗ extract the Iris features. In matching, the dissimilarity between a pair of Iris codes is defined by L1 norm. ∫, y 2πr ds ∂r r , x0 0 (1) 27 © 2010 ACEEE DOI: 01.ijsip.01.01.06 ACEEE International Journal on Signal and Image Processing Vol 1, No. 1, Jan 2010 Where I(x,y) is the Eye image, r is the radius, Gσ(r) is from 0 to 1 and θ is angle in the interval from 0 to 2 π . a Gaussian smoothing function, and s is the contour of The remapping of the Iris region from ( x, y ) Cartesian the circle given by (r, x0, y0). The operator searches for coordinates to the normalized non-concentric polar the circular path where there is maximum change in pixel representation is modeled as given by the Equations 2, 3 values, by varying the radius and centre x and y position and 4. of the circular contour. I ( x(r ,θ ), y (r ,θ )) = I (r ,θ ) (2) Eye Image With x(r ,θ ) = (1 − r )x p (θ ) + rxl (θ ) (3) Integro-Differential Operator y (r ,θ ) = (1 − r ) y p (θ ) + ryl (θ ) (4) Daugman’s Rubber sheet model where I ( x, y ) is the Iris image, ( x, y ) are the original Cartesian coordinates, (r ,θ ) are the corresponding normalized polar coordinates, and are the coordinates of Image Enhancement the pupil and iris boundaries along the θ direction as shown in Figure 3. Feature Extraction using IWT Database Hamming Distance Figure 3. Daugman’s Rubber Sheet Model Verified Iris The rubber sheet model takes into account Pupil dilation and size inconsistencies in order to produce a Figure 1. Block diagram of IWTHRS normalized representation with constant dimensions. The Iris region is modeled as a flexible rubber sheet anchored The IDO is applied iteratively with the amount of at the Iris boundary with the Pupil centre as the reference smoothing progressively reduced in order to attain precise point. The segmented Iris image is normalized to a size localization and also Eyelids are localized with the path 60 * 250. of contour integration changed from circular to an arc. The Integro-Differential can be seen as a variation of the C. Image Enhancement Hough Transform, as it makes use of first derivatives of In order to obtain best features for Iris verification, the image and performs a search to find geometric polar transformed image is enhanced using contrast- parameters. The IDO works with raw derivative limited adaptive histogram equalization . The results information and hence it does not suffer from the of image before and after enhancement are shown in threshold problems of Hough Transform. The segmented Figure 4. Iris image is shown in Figure 2. (a) (b) Figure 4. (a) Normalized Iris before enhancement. Figure 2. Segmented Iris with occluding Eyelids and Eyelashes made (b) Normalized Iris after enhancement. black D. Feature Extraction B. Daugman’s Rubber Sheet Model Feature extraction is the most important step in Iris The homogenous rubber sheet model devised by Verification. We use Haar Integer Wavelet Daugman remaps each point within the Iris region to a Transformation to extract the features from the pair of polar coordinates (r ,θ ) where r is in the interval normalized Iris image. The normalized Iris image of size 28 © 2010 ACEEE DOI: 01.ijsip.01.01.06 ACEEE International Journal on Signal and Image Processing Vol 1, No. 1, Jan 2010 60*250 is subjected to Integer Wavelet Transformation to totally different. If two patterns are derived from the get Approximation band, Horizontal band, Vertical band same Iris, the Hamming Distance between them is close and Diagonal band. The Horizontal Detail band obtained to zero, since they are highly correlated and the bits after the first level Integer Wavelet Transformation is should agree between the two Iris codes. However, further subjected to two levels of decomposition. The because of the presence of noise due to Eyelid and approximation band obtained after the third level Eyelashes occlusion, the Hamming Distance may vary up decomposition consists of the prominent features. The to 0.4 even for the same Iris images captured at different horizontal band is selected at the first two stages of instances. To increase the efficiency, we compare the Iris decomposition, because the normalized Iris image shows image under test with all the 7 images of each group and more details in the horizontal direction i e., angular the mean value of the 7 Hamming Distances is used to dimensions of the actual Iris image compared to the decide whether the Iris image under test belongs to the vertical direction i e., the radial dimension of the actual same group or not. If the average Hamming Distance Iris image. The two dimensional approximation band obtained is greater than 0.39 then the subject is rejected containing the prominent features is converted into a one and if the average Hamming Distance is lesser than 0.39 dimensional array and it is binarized. To binarize, we then the subject is accepted as genuine. equate all the positive features to 1 and the negative features to 0. This finally results in a feature vector of IV. ALGORITHM size 256 bits. The conceptual model for the three levels Table 1 shows the Human Identification by IWTHRS Integer Wavelet decomposition for feature extraction is algorithm in which the authenticity of the test Iris image shown in Figure 5. is verified. Problem definition: LL LL HL Consider an Eye image of a subject whose identity has to LH HH be verified. The objective is to LL i) Segment the Iris with minimum noise, ii) Normalize the Iris, iii) Generate a minimum length feature vector, which LH HH includes all the distinct features of the Iris, and iv) verify the authenticity of the subject. Assumptions: LH HH i) The Eye image is captured using IR photography ii) The Eye image is a gray-scale image of size 150* 200 TABLE 1. IWTHRS ALGORITHM Input : Test Eye image. Figure 5. Conceptual diagram for 3 levels 2D Integer Wavelet Decomposition. Output: Verified Iris. E. Matching i. Segment the Iris image using IDO ii. Normalize the segmented Iris image from Matching between the two Iris feature vectors is done Cartesian coordinates to the normalized using Hamming Distance. It is a measure of how many non-concentric polar representation of size bits are the same between two bit patterns. Using the 60*250 using Daugman’s rubber sheet Hamming Distance of two bit patterns, a decision is made model as to whether the two patterns were generated from iii. Enhance the image using contrast limited different Irises or from the same one. In comparing the bit adaptive histogram equalization patterns X and Y, the Hamming Distance HD, is defined iv. Apply Integer Wavelet Transformation to as the sum of disagreeing bits over N, the total number of the normalized Iris image bits in the feature vectors and is given by the Equation 5. v. Subject the horizontal detail band obtained 1 N in step 4 to two level IWT HD = ∑ X j ⊕ Yj N j =1 (5) vi. Convert the approximation band obtained in step 5 into single dimension vii. Binarize the one dimensional array Since an individual Iris region contains features with viii. Find the Hamming Distance between the high degrees of freedom, each Iris region produces a bit- binarized feature vectors obtained in step7 pattern which is independent to that produced by another with the corresponding feature vector in Iris. On the other hand, two Iris codes produced from the the database same Iris will be highly correlated. In ideal case, if two ix. If HD<0.39, the subject is accepted as bits patterns are completely independent, such as Iris genuine, else rejected templates generated from different Irises, the Hamming Distance between the two patterns is high. This occurs because independence implies the two bit patterns will be 29 © 2010 ACEEE DOI: 01.ijsip.01.01.06 ACEEE International Journal on Signal and Image Processing Vol 1, No. 1, Jan 2010 V. PERFORMANCE ANALYSIS We tested the IWTHRS model on the CASIA Iris image database–version 1.0, which contains 756 gray- scale Eye images with 108 unique Eyes or classes and 7 different images of each unique Eye. The algorithm is simulated on MATLAB 7.4 version. For the performance analysis, we considered 200 gray-scale Iris images out of available 756 Iris images. Table 2 gives the value of FAR, FRR and EER obtained with Hamming Distance threshold of 0.39 for IWT and DWT. It is observed that the value of FAR, FRR and EER are less in the case of Iris verification by IWTHRS compared that by Figure 6. Graph of FAR and FRR for DWTHRS DWTHRS. TABLE 2. COMPARISON OF FAR, FRR AND EER VALUES FOR IWTHRS AND DWTHRS. IWTHRS DWTHRS FAR 0.11 0.19 FRR 0.105 0.135 EER 0.107 0.165 Table 3 gives the average computation time for different steps viz., Segmentation, Normalization, Enhancement, Feature extraction and Matching involved in IWTHRS and DWTHRS. It is observed that the time Figure 7. Graph of FAR and FRR for IWTHRS required for feature extraction in case of IWTHRS is only 0.16 ms when compared to 0.49 ms for DWTHRS. Thus VI. CONCLUSION IWTHRS reduces the computation time for feature extraction by 66%. This shows that the proposed system A novel technique for Human Recognition using Iris can perform better in real time. verification has been proposed. The scheme uses the Integer Wavelet Transformation on the normalized Iris TABLE 3. AVERAGE COMPUTATION TIME OF THE IWTHRS AND image to extract the distinct features of the Iris. The DWTHRS MODELS. implementation of IWTHRS in place of DWTHRS has remarkably improved the computation speed and IWTHRS DWTHRS efficiency. Matching the test image with a set of seven Time % of Time % of images instead of only one image and finding the mean of (ms) total (ms) total the Hamming Distances for decision making further time time improves the efficiency of the algorithm. Improvements Segmentation 12.92 94.7 12.92 92.48 can further be made by including Iris de-noising 2 techniques into the algorithm. Normalization 0.39 2.88 0.39 2.79 Enhancement 0.08 0.58 0.08 0.57 ACKNOWLEDGMENT Feature 0.16 1.17 0.49 3.50 1 Extraction The author is thankful to KRJS management, the Matching 0.09 0.65 0.09 0.064 Principal, Vemana Institute of Technology, and the Total Time 13.64 100 13.97 100 Principal, UVCE for providing the infrastructural facilities to carry out the research work. Figures 6 and 7 show the graph of FAR and FRR obtained for different values of Hamming Distance REFERENCES threshold to compare the performance of Iris based  J .Daugman, “High Confidence Visual Recognition of Human Recognition System using DWT and IWT for Persons by a Test of Statistical Independence,” IEEE feature extraction. As Hamming Distance increases, the Transactions on Pattern Analysis and Machine value of FRR decreases whereas FAR increases. The Intelligence, vol.15, no. 11, pp. 1148–1161, November value of EER is a point where FAR is equal to FRR. 1993.  J. Daugman, “Statistical Richness of Visual Phase Information: Update on Recognizing Persons by Iris Patterns,” International Journal of Computer Vision, vol. 45, no. 1, pp. 25–38, 2001.  L. Ma, T. Tan, Y. Wang, and D. Zhang, “Personal Identification based on Iris Texture Analysis,” IEEE 30 © 2010 ACEEE DOI: 01.ijsip.01.01.06 ACEEE International Journal on Signal and Image Processing Vol 1, No. 1, Jan 2010 Transactions on Pattern Analysis and Machine Bangalore. He obtained his BE and ME in Electronics and Intelligence, vol. 25, no. 12, pp. 1519–1533, December Communication Engineering from University Visvesvaraya 2003. College of Engineering, Bangalore. He was awarded Ph.D. in  W. Kong and D. Zhang, “Detecting Eyelash and Reflection Computer Science and Engineering from Bangalore University. for Accurate Iris Segmentation,” International Journal of He has over 35 research publications in refereed International Pattern Recognition and Artificial Intelligence, pp. 1025– Journals and Conference Proceedings. His research interests 1034, 2003. include Image Processing, Biometrics, VLSI Signal Processing,  C. Chin, A. Jin, and D. Ling, “High Security Iris computer networks. Verification System based on Random Secret Integration,” Proceedings of International conference on Computer Vision and Image Understanding, vol. 2, pp. 169-177, May 2005. K R Venugopal is currently the Principal and Dean, Faculty of  Ya-Ping Huang, Si-Wel Luo and En-Yi Chen, “An Engineering, University Visvesvaraya College of Engineering, Efficient Iris Recognition System,” Proceedings of the Bangalore University, Bangalore. He obtained his Bachelor of First International Conference on Machine Learning and Engineering from University Visvesvaraya College of Cybernetics, pp. 450-454, November 2002. Engineering. He received his Masters  N. Schmid, M. Ketkar, H. Singh, and B. Cukic, degree in Computer Science and “Performance Analysis of Iris-based Identification System Automation from Indian Institute of at the Matching Score Level,” IEEE Transactions on Science, Bangalore. He was awarded Information Forensics and Security, vol. 1, no. 2, pp. 154- Ph.D. in Economics from Bangalore 168, 2006. University and Ph.D. in Computer Science  C. Fancourt, L. Bogoni, K. Hanna, Y. Guo, R. Wildes, N. from Indian Institute of Technology, Takahashi, and U. Jain, “Iris Recognition at a Distance,” Madras. He has a distinguished academic career and has degrees Proceedings of International Conference on Audio and in Electronics, Economics, Law, Business Finance, Public Video based Biometric Person Authentication, pp. 1–13, Relations, Communications, Industrial Relations, Computer 2005. Science and Journalism. He has authored 27 books on Computer  Karel Zuiderveld, “Contrast Limited Adaptive Histogram Science and Economics, which include Petrodollar and the Equalization,” Proceedings of the First International World Economy, C Aptitude, Mastering C, Microprocessor Conference on Visualization in Biomedical Computing, pp. Programming, Mastering C++ etc. He has been serving as the 337-345, May 1990. Professor and Chairman, Department of Computer Science and Engineering, University Visvesvaraya College of Engineering, Prashanth C R received the BE degree Bangalore University, Bangalore. During his three decades of in Electronics and the ME degree in service at UVCE he has over 200 research papers to his credit. Digital Communication from His research interests include computer networks, parallel and Bangalore University, Bangalore. He distributed systems, digital signal processing and data mining. is pursuing his Ph.D. in Computer Science and Engineering of Bangalore L M Patnaik is the Vice Chancellor, Defence Institute of University under the guidance of Dr. Advanced Technology (Deemed K. B. Raja, Assistant Professor, Department of Electronics and University), Pune, India. During the past Communication Engineering, University Visvesvaraya College 35 years of his service at the Indian of Engineering. He is currently an Assistant Professor, Dept. of Institute of Science, Bangalore, He has Electronics and Communication Engineering, Vemana Institute over 500 research publications in refereed of Technology, Bangalore. His research interests include International Journals and Conference Computer Vision, Pattern Recognition, Biometrics, and Proceedings. He is a Fellow of all the four Communication Engineering. He is a life member of Indian leading Science and Engineering Society for Technical Education, New Delhi. Academies in India; Fellow of the IEEE and the Academy of Science for the Developing World. He has received twenty Shashikumar D R is a Professor in the national and international awards; notable among them is the department of Computer Science and IEEE Technical Achievement Award for his significant Engineering, Cambridge Institute of contributions to high performance computing and soft Technology, Bangalore. He obtained his computing. His areas of research interest have been parallel and B.E. Degree in Electronics and distributed computing, mobile computing, CAD for VLSI Communications Engineering from circuits, soft computing, and computational neuroscience. Mysore University, Mysore and Masters degree in Electronics from Bangalore University, Bangalore. He is pursuing research in the area of Biometric applications. His area of interest is in the field of Digital Image Processing, Microprocessors, Embedded systems, Networks and Biometrics. He is life member of Indian Society for Technical Education, New Delhi. K B Raja is an Assistant Professor, Dept. of Electronics and Communication Engineering, University Visvesvaraya college of Engineering, Bangalore University, 31 © 2010 ACEEE DOI: 01.ijsip.01.01.06