International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 7July 2013

A New Approach for Iris Recognition

Hannath C M#1, Shreeja R*2
#1 #2

MTech student(CSE),MES College of Engineering,Kuttippuram,India Assistant Professor(CSE),MES College of Engineering,Kuttippuram,India

Abstract Biometric utilize something you are to authenticate identification. This might include fingerprints, retina pattern, iris, hand geometry, vein patterns, voice password or signature dynamics. Among the various techniques, iris recognition is regarded as the most reliable and accurate biometric recognition system. In this paper, we propose a new approach for iris recognition. The proposed method consists of 3 steps, iris segmentation, feature extraction and matching. All the images used in this paper were collected from CASIA and UBIRIS iris database. Experimental result shows that proposed method provides smaller EER, higher CRR and fast feature extraction. Keywords Contourlet transform, Gabor filter, Hamming distance ,Hough transform, Iris recognition.

recognition system is highly related to the proper design of its subsystems.

I. INTRODUCTION Biometrics is the automated recognition of individuals based on their behavioural and biological characteristics. A good biometric should be Universal, unique and permanent. There are two types of biometrics, biological/physiological biometrics and behavioural biometrics. Former includes God created characteristics possessed by individuals such as face, hand geometry, retina, iris etc and latter refers to characteristics acquired by the individual throughout his life time, such as voice, signature etc. The iris is an externally visible, yet protected organ whose unique epigenetic pattern remains stable throughout adult life. These characteristics make it very attractive for use as a biometric for identifying individuals. Fig. 1 shows an eye image along with iris region. The probabilty that two irises could produce exactly the same iris pattern is approximately 1 in 10 .(The population of the earth is around 10 ). Iris recognition system is composed of several subsystems. These subsystems include segmentation localizing the iris region in an eye image, normalization creating a rectangular block of iris pattern from the circular iris region to eliminate dimensional inconsistencies, feature encoding generating a template containing only the significant features of the iris region, matching and classification-measuring the similarity between two iris templates. The overall performance of an iris
78 10

Fig 1: Eye image

II. PREVIOUS WORKS Research in the area of iris recognition has been receiving considerable attention and a number of techniques and algorithms have been proposed over the last few years. Daugman[1] is the first one to give an algorithm for iris recognition. His algorithm is based on Iris Codes. For the preprocessing , step inner and outer boundaries of the iris are located.. Integro- differential operators are then used to detect the centre and diameter of the iris, then the pupil is also detected using the differential operators, Feature extraction algorithm uses the modified complex valued 2-D Gabor filters. For matching, Hamming Distance has been calculated by the use of simple Boolean Exclusive - OR operator and for the perfect match give the hamming distance equal to zero is obtained. The algorithm gives better accuracy but the time required for matching and feature extraction is very high. Wildes[2] has made use of an isotropic band-pass decomposition derived from application of Laplacian of Gaussian filters to the image data. Like Daugman , Wildes also used the first derivative of image intensity to find the location of edges corresponding to the borders of the iris. The Wildes system explicitly models the upper and lower eyelids

ISSN: 2231-2803

http://www.ijcttjournal.org

Page 2290

International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 7July 2013
with parabolic arcs whereas Daugman excludes the upper and the lower portions of the image.

Input Image
In Tans method [3] a bank of circular symmetric filters is used to capture local iris characteristics to form a fixed length feature vector. In iris matching, an efficient approach called nearest feature line (NFL) is used. Constraints are imposed on the original NFL method to improve performance.

Denoising
Iris recognition system developed by Li Ma [4] is characterized by local intensity variations. The sharp variation points of iris patterns are recorded as features. Using wavelet analysis, record the position of local sharp variation points in each intensity signal as features .Exclusive OR operation for matching. This paper describes an efficient algorithm for iris recognition by characterizing key local variations. The basic idea is that local sharp variation points, denoting the appearing or vanishing of an important image structure, are utilized to represent the characteristics of the iris. The whole procedure of feature extraction includes two steps: 1) a set of one-dimensional intensity signals is constructed to effectively characterize the most important information of the original two-dimensional image; 2) using a particular class of wavelets, a position sequence of local sharp variation points in such signals is recorded as features. Li [5] presented an iris recognition algorithm based on modified Log-Gabor filters. The algorithm is similar as the method proposed by Dogman in general procedure while modified Log-Gabor filters are adopted to extract the iris phase information instead of complex Gabor filters used in Daugmans method. The advantage of Log- Gabor filters over complex Gabor filters is the former are strictly band pass filters and the latter are not. The property of strictly band pass makes the Log-Gabor filters more suitable to extract the iris phase features regardless of the background brightness. III. PROPOSED DESIGN In any real time biometric system accuracy and recognition time are crucial parameters. All the methods mentioned in the literature survey use normalization step, ie the annular iris pattern is transformed into a polar coordinate system or is unwrapped into a rectangular block. Then, feature extraction attempts to extract the iris information from the normalized iris image to generate a feature vector. But it is a time consuming process. In our proposed design we avoid normalizatioin step by extracting features directly from segmented iris. Our proposed design is shown in fig.2.

Pupil Detection

Outer Boundary Detection

Segmented Image

Contourlet Transform

Contourlet Coefficients

Possibilistic Fuzzy Matching

Matched Image

Fig 2: Proposed Design

ISSN: 2231-2803

http://www.ijcttjournal.org

Page 2291

International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 7July 2013
A. Segmentation Based on Gradient Contours First of all, colour image can be converted to gray scale. Then, denoising can be done using wavelet transform. Compared with other part of eye , the pupil is darker. So,it can detected by finding minimum values in the eye image. Then gray scale image converted into binary. Morphological operations like erosion and cleaning can be done on the binary images. Centroid of minimum intensity values can be calculated. Travel radially from centroid to find partial derivatives of intensity gradients in both x and y directions. If the gradient is greater than a specified threshold mark that point. Repeat this process and find all the points and join these discrete points to find the curve. Traditional methods like Canny operator can be used for edge detection. After that we scan radially around the pupil where there is maximum change in the pixel value. Then we draw a circle according to the equation of circle. Find perimeter of the circle and we get the sclera-iris boundary. Steps of segmentation are shown in fig.3.

c) Outer boundary detection

d) Segmented Iris Fig 3.Illustration of segmentation

B. Feature Extraction a)
Original eye image

b) Pupil detection

Contourlets are constructed via filter banks and can be viewed as an extension of wavelets with directionality The contourlet transform [7][8] is a directional transform, which is capable of capturing contours and fine details in images. The contourlet expansion is composed of basis function oriented at various directions in multiple scales, with flexible aspect ratios. To capture smooth contours in images, the representation should contain basis functions with variety of shapes, in particular with different aspect ratios. Directionality and anisotropy are the important characteristics of contourlet transform. Directionality indicates that having basis function in many directions, only three direction in wavelet. The anisotropy property means the basis functions appear at various aspect ratios where as wavelets are separable functions and thus their aspect ratio is one. Due to this properties CT can efficiently handle 2D singularities, edges in an image.Flow graph of contourlet transform is shown in fig.4. It consists of two major stages: the sub-band decomposition and the directional transform.Block diagram is shown in fig.5. In contourlet transform, Laplacian pyramid does decomposition of images into subbands and then directional filter bank analyse each detail image.First,wavelet transform is used for edge detection and then local directional transform for contour segment detection.Directional filter bank is

ISSN: 2231-2803

http://www.ijcttjournal.org

Page 2292

International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 7July 2013
designed to capture high frequency components,Laplacian pyramid permits sub-band decomposition to avoid leaking of low frequencies into several directional sub-bands,thus directional information can be captured efficiently.
Fig 5: Contourlet coefficients

C. Matching Fuzzy matching determines the similarities between data sets, information, and facts. In PFM, both the position and local feature vector of each point are used to estimate the pose transformation and the point correspondence. The steps are given below:

1)

D1 I 1 ( x, y ) I 2 ( x, y )
I 1 ( x, y ) is the coefficient of trained images. I 2 ( x, y ) is the coefficient of test image.

(1)

Fig 4: Flow graph of contourlet transform

2)

D2 I 2 ( x, y ) ( I 2 ( x, y ) * rtn)

D1 D1 ( D2 * w1 )
rtn is the angle of rotation. 3) D3 I ( x, y ) ( I ( x, y ) * scl )
2 2

(2)

D1 D1 ( D3 * w2 )
scl is the scaling applied to image. w1 , w2 are the weightages. 4)

(3)

Compute minimum value of D1 and find matching Iris.

Fig 5: Block diagram of contourlet transform

D) Experimental Results The proposed iris recognition system was evaluated on the CASIA-IrisV1-Interval database [9] and on the UBIRIS.v2 [10] database . The CASIA database is a popular iris database and is widely adopted to evaluate the iris recognition system. The receiver operating characteristic (ROC) curve ,equal error rate (EER) and correct recognition rate (CRR) are used to evaluate the performance of the proposed method. The ROC curve is a false acceptance rate (FAR) versus false rejection rate (FRR) curve , which measures the accuracy of matching process and shows the overall performance of an algorithm. The FAR is the probability of accepting an imposter as an authorized subject and the FRR is the probability of an authorized subject being incorrectly rejected.The EER is the point where FAR and FRR are equal in value. The smaller the EER is, the better the algorithm. the algorithm.Correct recognition rate (CRR), is the ratio of the number of samples being correctly classified to the total number of test samples. ROC curve of the proposed method shown in fig.6. Comparison of the proposed method with traditional method are given in Table I
TABLE I
COMPARISON OF EER,AND CRR

There is a toolbox for contourlet transform available in the internet. Using this toolbox we can extract the features and we get the contourlet coefficients. Result is shown in fig.5.Then,we recomputed the mean and standard deviation of all images . This is done to reduce the error due to lighting conditions and background. Extract contourlet coefficients of all segmented iris.

ISSN: 2231-2803

http://www.ijcttjournal.org

Page 2293

International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 7July 2013
Proposed Design 0.13% 99.3% 177.981 sec

Methods Daugman Wildes Tan Keylocal Variation Method Log Gabor Filter Method Tsai et al Proposed System

EER 0.08 % 1.76 % 0.57 % 0.07 % 0.28 % 0.1482 % 0.13 %

CRR 100% 99.85% 100% 98.67 % 99.3 %

From the table it can be shown that our posed method using contourlet transform performs very well in presence of noise. IV. CONCLUSIONS In this paper, we improved Iris Recognition System using contourlet transform and possibilistic fuzzy matching. Proposed method achieves a higher accuracy because contourlet transform has capacity to capture comparatively richer directional information .Also fuzzy matching is used to compare two sets of feature points by using the information comprising the local features and the position of each point. Experimental results reveal that our algorithm achieves high performance even in noisy images. REFERENCES
[1] J. G. Daugman, High confidence visual recognition of persons by a test of statistical independence, IEEE Trans. Pattern Anal.Mach. Intell.,vol. 15, no. 11, pp. 11481161, Nov. 1993. [2] R. P.Wildes, Iris recognition: An emerging biometric technology, Proc. IEEE, vol. 85, no. 9, pp.13481363, Sep. 1997.

Fig.6: ROC curve of proposed method

We selected 150 iris images from 50 classes in which there are three samples in each class to evaluate the system performance.Also we include noisy input images(Gaussian,Speckle and Poisson).For comparison gabor filter method and contourlet transform based method is also implemented and compare our proposed method with method proposed by Tsai et al (using gabor filter).Comparison of EER ,CRR and feature extraction time are shown below.

[3] Li Ma, Yunhong Wang, Tieniu Tan,Iris recognition using circular symmetric filters, in Proc. 16th Int. Conf. Pattern Recognition, vol. II, 2002, pp. 414417. [4] LiMa, TieniuTan, YunhongWang, DexinZhang,"Efficient Iris Recognition by Characterizing Key Local Variations",IEEE Transactions On Image Processing, Vol. 13, No. 6, June 2004. [5] Peng Yao, Jun Li, Xueyi Ye, Zhenquan Zhuang,Bin Li,"Iris Recognition Algorithm Using Modified Log-Gabor Filters", 2006 IEEE. [6] Chung-Chih Tsai, Heng-Yi Lin, Jinshiuh Taur, and Chin-Wang Tao,"Iris Recognition Using Possibilistic Fuzzy Matching on Local Features"IEEE Transactions on systems, man, and cybernetics, [7] Minh N. Do and Martin Vetterli "The Contourlet Transform: An Efficient Directional Multiresolution Image Representation",IEEE Transactions on image processing, Dec. 2005,Volume: 14 , Issue: 12, Page(s): 2091 - 2106.

TABLE III
COMPARISON OF EER,CRR AND FEATURE EXTRACTION TIME

Method EER Tsai et als system 0.1482%

Parameters CRR Feature Extraction Time 98.67 % 310.73 sec

[8] Amir Azizi and Hamid Reza Pourreza,"A New Method for Iris Recognition Based on Contourlet Transform and Non Linear Approximation Coefficients",springer,pp. 307316, 2009. [9] CASIA-IrisV3. http://www.cbsr.ia.ac.cn/IrisDatabase.htm [Online]. Available:

[10] H. Proena and L. A. Alexandre, UBIRIS: A noisy iris image database,in Proc. Int. Conf. Image Anal. Process., 2005, vol. 1, pp. 970977.

ISSN: 2231-2803

http://www.ijcttjournal.org

Page 2294