Agent-based image iris segmentation and multiple views boundary refining

摘要: The paper presents two different methods to deal with the problem of iris segmentation: an agent-based method capable localize center pupil and a process boundaries by multiple views approach. In first method, agent corresponds coordinates specific point analysis in input image. A population agents is deployed image, then, each collects local information concerning intensity patterns visible its region interest. By iterations, changes position accordingly properties, moving towards estimation center. If no available present interest, will move itself along random walk. After few tends spread then concentrate inner portion pupil. Once has been located, outer are refined approach based on analysis. This starts set points that can be considered as approximation For point, detailed computed, final description obtained merging all descriptions. were tested using CASIA v.3 UBIRIS v.2 images. Experiments show proposed approaches feasible, also eye images taken noisy or non-ideal conditions, achieving total error segmentation accuracy up 97%.