Iris Recognition Basing on Multi-resolution Analysis

Multi-resolution decomposition is very effective for analyzing and extracting the information content of images. According to this idea, we apply this theory to iris recognition system. In the iris localization part, in order to avoid the influence of eyelash, we detect the edge points of iris through the detail signals of some rows at appropriate resolutions. Then, we set two circles to these edge points to obtain the exact boundary of iris. In the feature extracting part, we select out some rows of the normalized iris image after enhancing as the feature vector and only extract the approximation signals of these vectors to construct the final feature vector for matching. The greatest merit of this algorithm is its stability, for the results are seldom affected by lash occlusion problem. From our experiments on 2400 iris images of CASIA (the Institute of Automation, Chinese Academy of Sciences), the validity of this algorithm has been testified