The Viterbi algorithm at different resolutions for enhanced iris segmentation

We propose in this paper a novel method for iris segmentation. In order to retrieve iris contours, the Viterbi algorithm is applied on the gradient map of images processed by Anisotropic Smoothing. The Viterbi algorithm is exploited at two resolutions: at a high resolution, it allows finding precise contours, while at a low resolution, coarse contours that improve the accuracy of normalization circles are retrieved. We tested the method on several databases (ICE 2005, NDIRIS-0405, Casia-V3-Interval). Our extensive experiments on such databases lead to performance at the state of the art, while the proposed method does not require a refined parameter adaptation to the various degradations encountered.