Iris recognition in nonideal situations

Abstract

Most of the state-of-the-art iris recognition algorithms focus on processing and recognition of the ideal iris images which are captured in a controlled environment. In this paper, we process the nonideal iris images which are acquired in an unconstrained situation and are affected severely by gaze deviation, eyelids and eyelashes occlusion, non uniform intensity, motion blur, reflections, etc. To segment the nonideal iris images accurately, we deploy a variational level set based curve evolution scheme, which uses significantly larger time step for numerically solving the evolution partial differential equation (PDE), and therefore, speeds up the curve evolution process drastically. Genetic Algorithms (GAs) are deployed to select the subset of informative features by combining the valuable outcomes from the multiple feature selection criteria without compromising the recognition accuracy. The verification performance of the proposed scheme is validated using three nonideal iris datasets, namely, UBIRIS Version 2, ICE 2005, and WVU datasets. © 2009 Springer Berlin Heidelberg.