Iris image selection and localization based on analysis of specular reflection

We present a method for selecting good quality iris images from a sequence based on the position and quality of the specular reflection relative to the pupil. Using a simple low complexity yet robust technique, fast image selection is achieved along with immediate iris localization. Initial screening is achieved by accepting images with appropriate specular reflection intensity, width, height and area. Finer selection is then based on graylevel changes between light-pupil and pupil-iris regions. The method has been evaluated on numerous problem images and all were correctly rejected. Connected-component algorithms are used to find exact pupil boundaries and iris edges are modeled by ellipses based on left-right segmented boundaries. On a test set of 400 classes consisting of over 60000 images, no false acceptances are noted.