Iris Recognition Using Visible Images Based on the Fusion of Daugman's Approach and Hough Transform

Abstract

In the paper, we proposed a novel architecture for Iris Recognition. Contrary to the conventional approaches, where iris is obtained using NIR images, iris recognition is performed using visible images. In the proposed methodology, iris localization is achieved using a mask generated using Hue and saturation channels. Then the localized iris region is converted to binary mask using a threshold. After the binary mask is generated, the inner circle of the iris is obtained using the fusion of a contour-based approach and Hough transform. Once the inner circle of the iris is computed, the outer circle of the iris is estimated using Daugman's approach. After the inner and outer circles of the iris are segmented normalization is achieved by converting the polar coordinates to Cartesian coordinates and features are extracted. In the proposed architecture, we have done a comparative analysis using LBPH features and Zernike features and two classifiers, random forest and Support Vector Machines. During the recognition module, the feature is extracted from the test image and compared against the existing database.