A fast and accurate pupil localization method using gray gradient differential and curve fitting

Abstract

Pupil localization plays a key role in recognizing the biological characteristics of iris that is self-evident. In recent years, the most common procedures to localize the pupil have been based on the edge detector and circle finder, such as integro-differential operator and Hough transform. However, the circle finder overemphasizes geometric characteristics, which reduces the accuracy and real-time performance under complex conditions. In this chapter, a new threshold method using discrete gray gradient differentials based on the unique features of gaps in the gray gradients of pupil boundaries is proposed to binarization first. We then highlight the geometric characteristics of the pupil by adopting the strategy of combining outline filling with curve fitting to locate pupil boundaries. Compared to the Wildes system, the proposed method is feasible, fast, accurate, and stable.