A Robust IRIS Segmentation Procedure for Unconstrained Subject Presentation

Iris as a biometric, is the most reliable with respect to performance. However, this reliability is a function of the ideality of the data, therefore a robust segmentation algorithm is required to handle non-ideal data. In this paper, a segmentation methodology is proposed that utilizes shape, intensity, and location information that is intrinsic to the pupil/iris. The virtue of this methodology lies in its capability to reliably segment non-ideal imagery that is simultaneously affected with such factors as specular reflection, blur, lighting variation, and off-angle images. We demonstrate the robustness of our segmentation methodology by evaluating ideal and non-ideal datasets, namely CASIA, Iris Challenge Evaluation (ICE) data, WVU, and WVU Off-angle. Furthermore, we compare our performance to that of Camus and Wildes, and Libor Masek's algorithms. We demonstrate an increase in segmentation performance of 7.02%, 8.16%, 20.84%, 26.61%, over the former mentioned algorithms when evaluating these datasets, respectively.