Iris-Based Personal Authentication Using a Normalized Directional Energy Feature

Abstract

In iris-based biometric systems, iris images acquired by a video or CCD camera generally have a lot of contrast or brightness differences in an image or between images due to the different extent of camera focusing or illumination conditions. To extract the discriminatory iris features robust to such differences, this paper presents a new normalization scheme of the directional energy that will be used as the iris feature. The proposed method first performs band-pass filtering on the input iris image to reduce the effect of high frequency noise and DC energy difference, then decomposes the image into several directional subband outputs using a directional filter bank (DFB). The directional energy values of the iris pattern are extracted from the decomposed subband outputs on a block-by-block basis and then are normalized with respect to the total block energy. Matching is performed by finding the Euclidean distance between the input and enrolled template feature vector. Experimental results show that the proposed method is robust to changes in illumination or contrast and the overall feature extraction and matching procedures are effective.