Iris Localization via Pulling and Pushing

Iris localization is a critical module in iris recognition because it defines the inner and outer boundaries of iris region used for feature analysis. State-of-the-art iris localization methods need to implement a brute-force search of the large parameter space, which is time-consuming and sensitive to noises. This paper proposes a novel iris localization method based on a spring force-driven iteration scheme. First, the coarse localization of pupil is obtained by an AdaBoost-based iris detection method. Then the radial edge points are detected in polar coordinate, which contribute a force on the circle center based on Hooke´s law. Finally the center and radius of pupil and iris are refined according to the composition of forces from all points. After 3-4 iterations, the precise location of iris could be found. Experimental results show that our method is faster and more accurate than state-of-the-art iris localization methods