Accurate pupil boundary detection using angular integral projection and bezier curves
Ghassan J. Mohammed Aladool, Ann A. Jarjes Alkazzaz · 2018
Iris localization is an important key step in the whole iris recognition system, it detects both iris inner and outer boundaries. In this paper, an algorithm for accurate pupil boundary detection is proposed, it detects pupil boundary points using the angular integral projection function (AIPF) and then fits a set of Bezier curves for these boundary points. So far, many existing methods for iris localization cannot detect pupil boundary accurately, but instead they consider the pupil as circular region and fit a circle to its boundary, which of course degrades the performance of matching process. The proposed algorithm works in several steps. First, the approximate pupil center is found by calculating the center of mass for the binarized iris image. Then, a set of pupil boundary points is detected using AIPF. After that, two intermediate Bezier's control points are obtained for each subset of boundary points using the least square method. Finally a cubic Bezier curve is fitted to each subset of boundary points and control points. Experimental results on 2655 iris images from CASIA V3.0 demonstrate the efficiency and accuracy of the proposed algorithm.