FISHER LINEAR DISCRIMINANT ANALYSIS TECHNIQUE FOR IRIS RECOGNITION

Abstract

With an increasing emphasis on security, automated personal identification and verification based on biometrics has been receiving extensive attention over the past decade. Iris recognition, as an emerging biometric recognition approach, is becoming a very active topic in both research and practical applications. The motivation for this endeavor stems from the observation that the human iris provides a particularly interesting structure on which to base a technology for noninvasive biometric assessment. In particular, the biomedical literature suggests that irises are as distinct as fingerprints or patterns of retinal blood vessels. Further, since the iris is an overt body, its appearance is amenable to remote examination with the aid of a machine vision system. In general, a typical iris recognition system includes iris imaging, iris liveness detection, and recognition.

This research work deals with the implementation of iris recognition using Fisher Linear Discriminant Analysis and PCA method on iris images. Our projection method is based on Fisher's Linear Discriminant and produces well separated classes in a low-dimensional subspace. The proposed system contains four parts i.e. preprocessing, segmentation, feature extraction and matching. The preprocessing part further contains pupil localization, image refinement, iris localization and normalization processes. First of all, we acquire the grayscale eye image. In second step, the radius and centre coordinates of the pupil is calculated. For this purpose, apply the thresholding with the value zero in order to make the pupil appear alone. we calculate the longest black row horizontally, and the longest column vertically, by counting the number of black points in each row and in each column as in figure, then find out the coordinate of the center easily and also the radius. 4 Fisher Linear Discriminant Analysis Technique for Iris Recognition In third step, we find the threshold value (T) Using Basic Global Thresholding algorithm. Implement the thresholding with T from previous algorithm on original image. In fourth step, complement of the image is taken and applied the median filter for image refinement. In the fifth step, applying sobel edge detection on the refined image to detect the edges in the image. In the sixth step, after the edge detection of the image, we apply the Hough transform circle algorithm to find the centre coordinates and radius of the iris. The iris is the area between outer circle and inner circle. We got the pixels of this area and colored the background with black. So we cut the iris region in donut like shape. Then we whiten the non-iris region and display the iris region in the image only. Typical next step is to “normalize” the iris to a rectangular representation. For every pixel in the iris, an equivalent position is found out on polar axes. The normalized image was then interpolated into the size of the original image. In the next step, we create the database of the normalized images. Read gray normalized images one by one and convert it into 1-D image and save it into associative memory. In the 11th step, feature extraction process takes over in which firstly build the projection matrix by projecting central image onto eigen space by using Principal Component Analysis and this matrix becomes an input to Fisher Linear Discriminant Analysis Method to build Fisher projection matrix by projecting images onto Fisher linear space. The output of this process is encoded in the form of feature template. Then this feature template of the input iris image is stored in the database. In the final 12th step, when the iris identification is required then the test image is compared with the feature templates that are stored in database and calculate the Euclidean distance between the test images and stored images. The image with the 5 Fisher Linear Discriminant Analysis Technique for Iris Recognition minimum Euclidean distance compares with threshold value that describes the identification of test image either it exists in the database or not. The proposed algorithm was tested on CASIA database. The empirical results provide the accuracy of 97% with time delay of 0.01064 sec per image. The comparison of the proposed technique with other well known techniques is provided in the thesis both with respect to time and performance in the form of graphs. The robustness and time efficiency of the proposed algorithm makes it perfect system for real time applications.