Method of Optimal Circular Path for Iris Template Matching

Abstract

A modification of the conventional method for iris template matching is proposed. The normalized iris images to be compared are divided into several sectors, which are then shifted in the angular and radial directions to find the best alignment. The values of these shifts are determined by finding the optimal path in the distance array. This approach is algorithmically and computationally simpler than its well-known analogs and yields similar or better results in terms of recognition accuracy. In addition, the problem of selecting the optimal parameters for the wavelets that form the iris pattern is investigated. It is shown that improving the accuracy of alignment makes it possible to use more informative high-frequency wavelets. Numerical experiments are carried out on the ICE2005 and CASIA open image databases.