الفهرس | Only 14 pages are availabe for public view |
Abstract Reliable automatic recognition of persons has long been an attrac- tive goal. Much of the work in the eld of biometrics has focused on identication applications. Iris biometric systems can function as extremely reliable means of personal identication due to several rea- sons. The iris is extremely data-rich characteristic trait with unique non-genetic patterns. Moreover, these patterns are stable throughout the adult life and are physically protected by the cornea that does not inhibit external viewability. The rst -yet critical- step in iris recognition is to isolate the iris region in the digital image of the eye. This process is called iris lo- calization and segmentation, and is the focus of this thesis. Di erent image processing techniques that involve numerical optimization are used for iris segmentation and localization. This thesis reviews and implements di erent iris localization and segmentation techniques and compares between them. The iris region can be approximated as two circles representing iris/pupil boundary and the iris/sclera boundary. Circular Hough Transform and Integro- Di erential operator assume that the iris/pupil and iris/sclera bound- aries are circular. Active contours, a newly utilized technique in iris segmentation, attempt to nd the boundaries assuming that they are I arbitrary by iteratively evolving curves until they lock on nearby edges. The di erent iris segmentation techniques are implemented and tested on database of irises provided by the Chinese Academy of Sci- ences Institute of Automation (CASIA). II |