Localization, Segmentation, Feature Extraction, Classification And Identification Of Imperfect Irises

For national security, social security, finance safety and flight safety, a more convenient and reliable verification and identification method is required. Development and progress of biometrics technology can provide an ideal solution. Iris recognition is an aspect of biometrics and has been successfully applied to person verification and identification in some fields. At present, the iris recognition system requires user's cooperation with iris image acquisition set. However, precise image acquisition is not always possible and this will bring the difficulties to the subsequent processing. In this dissertation, this problem is studied in details. The main contributions are as follows:(1) Due to unideal real-time character of iris boundary localization using Hough transform and integral-differential method, the fast boundary localization algorithms are proposed based on pupil position estimation. In order to accurately segment the iris, we present a combined-line approach for accurate eyelid localization and adaptive threshold algorithms for detection of eyelashes and light spots. Simulation results show that the effective iris region can be segmented.(2) In general, the size and resolution of iris images will be different and it is difficult to process, thus we will transform the annular iris region into the rectangle one, and normalize the iris region into the same resolution based on coherence at a rough scale. The classification results show that normalized region still has enough information to distinguish different iris patterns.(3) We present a simple spatial domain zero-crossing detection algorithm to...