Non-parametric Iris Localization Using Pupil's Uniform Intensities and Adaptive Masking
Ritesh Vyas, Tirupathiraju Kanumuri, Gyanendra Sheoran · 2017
Iris localization is one of the vigorous components of any iris recognition system. It deals with the separation of annular iris from the acquired eye image. Accuracy of the iris segmentation module directly affects the overall system accuracy. In order to localize the pupillary boundary, this work utilizes local binary pattern (LBP) to exploit the uniform intensities present in the pupil region, to detect its boundaries. LBP aids in reducing the adverse effects of eyelashes in pupil localization. Moreover, an adaptive mask is also developed for localizing the limbus boundary. This mask provides a mean to combat with the varying sizes of iris due to variation in illumination. Outcomes of experiments performed with two benchmark iris databases (i.e. CASIA-IrisV1 and IITD iris database) support the efficacy of the proposed approach.