ISSN 2249-6343 International Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume 3, Issue 6, December 2013 1 Person Identification Technique Using Human Iris Recognition S.M. Ali Remote Sensing Research Unit, College of Science, University of Baghdad, Iraq, Baghdad, Al-Jaderyia Abstract— A new iris recognition and person identification technique is introduced. The method is based on tracing the Eye image boundary, using the Marr- Hildreth edge technique. The Eye's pupil boundary is identified, using the contour follower which is based on the chain coding method. The coordinates of the pupil's center is then recognized as the center of figure of the pupil boundary chained points. The pupil radius (i.e. the iris inner radius) is determined as the average distance between the pupil center and the chained points on the pupil boundary. An image of 128128 pixels size surrounding the pupil center is then extracted which involve all the required necessary information for the identification process. The extracted raster form images then transformed from the Euclidian representation into polar representation (i.e. r- coordinates), all having the same size 36060 pixels (i.e. rows number for inner iris radius ≤ r≤ outer iris radius, and columns number for 0 o ≤  ≤ 360 o ). The transformed polar represented images are, then, inserted (each as a column vector) in an array, their average is determined and subtracted from each. The covariance of the average subtracted array then computed, the eigenvectors of the covariance matrix are, then, calculated and used to identify the tested eye by comparing its eigenvector with those already preserved as Database, using the minimum mean- square criterion. Index Terms—iris recognition, iris identification, iris verification, edge detection, contour following I. INTRODUCTION Iris is part of eye between eyelids and surrounding. For long time the high confidence authentication of human identity was demanding by many highly securitizing societies, the biometric person identification technique based on the pattern of the human iris has been found as a well suited to be applied for many access control systems requiring a high level of security. Identification based on iris patterns is one of the most reliable and stable way among all available biometrics [1]–[5]. The work presented in this research introduces a new iris recognition and identification system, based on performing several image processing techniques (i.e. edge-detection, contour following, image transformation, image normalization, image extraction, and matching operations), as will be illustrated below. II. MARR-HILDRETH EDGE DETECTION TECHNIQUE The creation of an edge image can be considered as a process of transforming the grey scale image into binary- valued image. This transformation is recognized as to be performed relatively fast by implementing a threshold differentiation operators (i.e. Robert, Prewitt, Sobel, etc), [6]. The problem with these techniques is existed in its threshold dependence; i.e. many spurious edges are detected with certain low-valued threshold, or can give discontinuous edges if higher threshold is used to eliminate the spurious edges. Too many edge detection methods have been proposed in the literature [7]–[9]. The Marr-Hildreth suggested an edge method as being a model of human visual processing, which produce thin and connected boundaries [10], [11]. The procedures involved in this edge method can be summarized by the following points. 1) Since detecting edges in an image with large scale of intensity changes requiring some form of smoothing, an optimal smoothing can be performed by convoluting the discrete image array f(x ,y), with a 2D-Gaussian operator G(x, y); i.e. ) , ( ) , ( ) , ( y x f y x G y x c   (1) Where: c(x, y) represents the smoothed image, * denotes spatial convolution process, and G(x, y) is Gaussian function, given by: ) 2 2 ( 2 2 2 2 2 1 ) , ( y x y x y x e y x G        (2) Under conditions of circular symmetry, 2 2 2 and r y x y x      , then G(x, y) becomes: 2 2 2 2 2 1 ) (   r e r G   (3) 2) The edge points can then be detected by searching for the zero-crossings in the 2 nd directional derivative of the intensity function of the image (i.e. Laplacian operator 2  ). The Laplacian-Gaussian operator can be combined into one operator " G 2  " and called the Marr-Hildreth operator, given by: 2 2 2 2 2 4 2 ]. 2 1 [ 1 ) (    r e r r G     (4) 3) Candidate edge points are then located by scanning the image, pixel-by-pixel, looking for zero-crossing signals.