Abstract
Iris segmentation plays an important role in an accurate iris recognition system. In this paper, we contribute to a more accurate iris segmentation process, hence, a more precise identification tool. We introduce a novel iris segmentation approach based on mixture of probability distributions modeling and an extended Expectation Maximization based on the (EM) algorithm. A segmentation approach is presented by exploring different techniques. Our analysis takes care of segmenting the eye image using the Markovian case and the independent case. In each case, the approach uses different mixture of probability distributions. While providing comparable results, our update algorithm gives better segmentation results based on our learning system. The accuracy of our algorithm is proved by Kulback-Leibler divergence and mean squared error computations. The statistical significance of our results are evaluated using Anova Test. The proposed method was applied on the CASIA testing database by Chinese Academy of Sciences Institute of Automation-Iris-Twins. The results showed, for both cases, the accuracy of the proposed algorithm verses the classical mixture normal distribution.