Efficient iris recognition by computing discriminable textons

This paper describes an efficient algorithm for iris recognition by computing the discriminable textons. The basic idea is that texture discrimination is based on first-order differences in geometric and luminance attributes of texture elements, called 'textons'. The whole procedure of feature extraction includes two steps. First a map of dark blob pixels are computed by convolving with a Gaussian filter followed by a Laplasian differential operator and then combining morphological operations with a two-threshold method, the most important blobs of iris are segmented on the basis of local shape into small compact and thin elongated components. Iris matching is implemented by calculating the hamming distances between two binary code sequences of irises. Experimental results indicate that the algorithm is successful in recognizing the different iris pattern especially when the iris images are not occluded by eyelids and eyelashes.