EM-Based Nonideal Iris Boundary Localization
Hong Lin Wan, Bao Sheng Li, Min Han, Deng Wang Li · Advanced materials research · 2012
Nonideal iris segmentation is a great challenge in iris recognition, and many researchers have stressed this problem. Since critical step of segmentation is localizing iris center and detecting interior/outer boundaries, we presents a novel method based on EM algorithm to deal with it. EM algorithm is capable of automatic threshold, therefore candidate pupil can be obtained and followed by an innovated fast iris center searching by using strings equilibrium scheme. We also give the region-based outer boundary localization with implementation of order statistical filters (OSF). Experiments demonstrate a high correct segmentation ratio (CSR) of more than 98% has been achieved when using CASIA-IrisV3 Interval and CASIA-IrisV3 Lamp databases.