Iris recognition by new local invariant feature descriptor

This paper presents a novel iris recognition based on windowed gray difference histogram of subregions. The keypoints are detected by scale-space extrema of difference-of-Gaussian function from the normalized iris images. For the subregions with keypoints, they are divided into blocks and gray difference histogram of each block is computed by sliding the window template. The feature vector of a subregion can be expressed by concatenating histograms of all blocks within the subregion. Then the histogram matching distance assesses the similarity between two iris images. Experimental results on CASIA Iris images show that the performance of the proposed method is encouraging and comparable to the state-of-the-art iris recognition methods. Copyright © 2013 Binary Information Press.