CDCF-based iris feature extraction and matching

Having investigated many public iris algorithms, we find they can be improved by utilizing iris feature sufficiently. In this paper we first analyze the properties of normalized iris image, and then propose a thought that do not optimize the accuracy of single matched pair but focus on coarse matched pairs filtering, due to precise matched pairs have their inherent positional relationship. Experiments show the significant improvements of the algorithm proposed, including compact storage requirement, fast matching speed, high accuracy and robustness.