Heterogeneous Iris Segmentation Based on Active Contour Model and Prior Noise Characteristics

Abstract

Recently, iris segmentation has attracted increasing attentions due to its importance for iris recognition systems. A lot of efforts have been made to design iris segmentation methods; however, heterogeneous iris segmentation remains challenged. Therefore, this paper presents a heterogeneous iris segmentation method based on active contour model and prior noise characteristics. By evaluating prior noise characteristics, the proposed method first classifies heterogeneous iris images into two types. Following this, it integrates merits of the localizing region-based active contours model and Hough transform to localize the interior and exterior boundaries of iris regions with a divide-and-conquer pattern. Extensive experiments demonstrate the proposed method overcomes the disturbances caused by occlusions, reflection, and texture, improving the accuracy and success rate of heterogeneous iris segmentation.