Real shape inner iris boundary segmentation using active contour without edges

Research and development in biometric recognition systems must respond to the industrial and experimental reported problems. When considering the iris recognition system, segmentation of the iris is considered as its most difficult step. Classic methodologies determine the inner and outer borders of the iris region using circular or elliptical boundary. These border shapes succeed to delimit correctly the iris region for ideal iris images it fails with non ideal images. In this paper we propose a method to detect the inner iris boundary in its real shape whatever its geometry. First circular Hough transform is used to define the pupil's region of interest, and, second active contour without edges is used to exactly segment the inner iris boundary. The method is tested on CASIA-V3 and UBIRIS databases and results are compared to Wildes' model. Visible improvements show the efficiency of the proposed method to adapt to the inner iris boundary real shape.