Quality-based dynamic threshold for iris matching

Current iris recognition systems usually regard poor quality iris images useless since defocused or partially occluded iris images may cause false acceptance. However, such a strategy may lose an opportunity to correctly report a genuine match with poor-quality samples. This paper proposes an adaptive iris matching method to improve the throughput of iris recognition systems. The core idea of the method is to dynamically adjust the decision threshold of iris matching module based on the quality measure of input iris image. So that the poor quality iris images also have a chance to match template database under the controlled false accept rate. Experiment results on the real system demonstrate the effectiveness of the proposed method and the recognition time is expected to be greatly reduced.