The Hermite Transform: An Alternative Image Representation Model for Iris Recognition

Alfonso Estudillo-Romero, Boris Escalante-Ramírez

Abstract: In this work we propose an alternative image representation model to efficiently characterize iris textures based on the Hermite transform. The Hermite transform can simulate some properties of the mammalian visual system and it is founded on a well established mathematical framework. These properties are used to extract the most important information of the iris textures. The results show that the Hermite transform is able to characterize iris textures as well as the Gabor model, with the advantage on the second that the discrete analysis filters in the Hermite transform are given by the Krawtchouk polynomials and, it is not needed to compute the filter coefficients by means of optimization methods, nor to suppress the zero mean (d.c. response). The proposed iris recognition system achieved an overall performance of 97.34% and a Correct Access Rate (CAR) of 90.29% when the False Access Rate (FAR) was closed to zero.