Block based texture analysis for iris classification and matching

The goal of this paper is to analyze the texture of irides and determine if they can be quantitatively measured and assigned into multiple categories. Such an exercise would ensure that irides, like fingerprints, can be partitioned into multiple classes thereby allowing for faster retrieval of identities in large scale biometric systems. In order to facilitate this, a set of 68 statistical features is extracted from the iris texture. These features correspond to the high frequency information associated with anatomical structures in the iris such as crypts, furrows and pigment spots. The statistical features extracted from different blocks in the iris are fused at the feature level and decision level. Experimental analysis using the UPOL database indicates the efficacy of the proposed scheme in (a) clustering iris texture, and (b) assigning an input iris to the correct cluster based on its textural content. The feasibility of using blocks of iris to perform partial iris matching is also investigated.