A novel iris segmentation method based on multi-line

Abstract

In the iris authentication system, the existence of eyelids and eyelashes noises might generate negative effect on pattern analysis. This paper proposes a novel and accurate iris segmentation method which adequately considers the edges and statistical feature to detect the eyelids and eyelashes noises in the captured iris images. First, multi-scale edge detection is used to get the iris coarse locations, and rank filter is employed to smooth images for determining a more accurate searching area of eyelids. Second, morphological operations as well as line Hough transform are presented to reserve the available edge points for multi-line eyelids fitting. Specially, the adaptive average intensity of individual iris image based on region of interest (ROI) is educed to get the statistical threshold for eyelashes detection. Experimental results indicate that the proposed method can effectively remove the occlusion caused by eyelids and eyelashes, and increase the amount of information (AOI) of segmented iris and improve the iris location accuracy. © Springer-Verlag Berlin Heidelberg 2013.