Indexing iris images

Abstract

Given a query iris image, the goal of indexing is to identify and retrieve a small subset of candidate irides from the database in order to determine a possible match. This can significantly improve the response time of iris recognition systems operating in the identification mode. In this work, we analyze two different approaches to iris indexing. The first technique is based on the analysis of IrisCodes (post-encoding indexing); the second technique is based on the analysis of features extracted from the iris texture (pre-encoding indexing). Experiments on a subset of the publicly available CASIA-IrisV3 database compare the two approaches and illustrate the potential of the proposed indexing methods for large scale iris identification.