A Robust and Accurate Segmentation of Iris Images Using Optimal Partitioning

An effective and accurate identification of human individuals from their iris features is largely dependent on proper segmentation of the iris and the pupil features from camera images. Most modern segmentation schemes exploit the circular geometry of the iris to fit a circle or an ellipse to an edge map of the iris. In this paper, we present a new method for automatically localizing and segmenting iris features by optimal partitioning using the relative distribution of gray-level intensities across an image. First, the eye images are unrolled after detecting the center of the pupil from the image local minima. For each radial sample, segments corresponding to regions that are statistically different are computed using dynamic programming applied to a Poisson-based cost function. The results are a set of change points marking the edges of different features including those of the pupil and the iris. The radius of the pupil and the iris are then obtained by searching for the best fit of two lines connecting the detected edge points. The proposed method is superior to other methods in that artifacts such as excessive or weak illumination, blurring and occlusion by eyelids do not interfere with the segmentation process. Moreover, our algorithm is also robust and accurate even in the presence of eyewear such as glasses. Applying this method to 122 images revealed a 98% segmentation accuracy. The algorithm has been shown to be effective in images with large field of view containing other facial features