Iris categorization with texton representation

A key concern with iris recognition systems is the time required to reliably find a test sample's match in a large database of subjects. This work considers methods for categorizing irises within a database, so that a search for a match to a test sample can be focused on the test sample's category. This work uses texton learning to reduce the representation of the images and then clusters the images with the unsupervised k-means technique. Success of the system is assessed as its ability to consistently classify images from the same subject. This work includes experiments to determine the optimal number of textons and image clusters. It also investigates different accuracy metrics and analyzes the potential time saving impacts for finding a database match.