Hausdorff Distance for Iris Recognition

Abstract

Iris is a promising biometric due to its high reliability and stability. In this paper, a novel iris recognition technique based on Hausdorff distance is proposed. A modified partial Hausdorff distance (a dissimilarity measure) is computed directly between the normalized iris images for comparison and no feature is extracted explicitly. The Hausdorff distance-based iris recognition system is expected to perform well in the case of severe occlusion by eyelids due to the partialness in the measure. Besides, the modified measure is insensitive to lighting conditions. Experimental results on the CASIA database show that the performance of the proposed recognition system is encouraging and comparable to the iris recognition algorithms found in the current literature.