Iris Segmentation Based on an Adaptive Initial Contour and Partly-Normalization

Abstract

Active contour is accurate for iris segmentation on the non-ideal and noisy iris images. However, understanding on how active contour reacts to the motion blur or blurry iris images is presently unclear and remains a major challenge in iris segmentation perspective. Moreover, studies on the initial contour position in the blurry iris images are infrequently reported and need further clarification. In addition, convergence or evolution speed is still a major drawback for active contour as it moves through the boundaries in the iris images. Based on the above issues, the experiment is conducted to obtain an accurate and fast iris segmentation algorithm for the blurry iris images. The initial contour is also investigated to clarify its positioning for the blurry iris segmentation. To achieve these objectives, the Wiener filter is used for pre-processing. Next, the morphological closing is applied to eliminate reflections. Then, the adaptive Chan-Vese active contour (ACVAC) algorithm is designed from the adaptive initial contour (AIC), δ and stopping function. Finally, the partly-normalization is designed where only prominent iris features near to the inner iris boundary are selected for normalization and feature extraction. The experimental results show that the proposed algorithm achieves the highest segmentation accuracy and the fastest computational time than the other active contour-based methods. The accurate initial contour position in the blurry iris images is clearly clarified. This shows that the proposed method is accurate for iris segmentation on the blurry iris images.