A New Method for Iris Recognition using Gray-Level Coccurence Matrix

In this paper we present a new method for iris texture recognition for the purpose of human identification using statistical analysis of gray-level distribution. Many studies have been aimed at extracting iris features that are unique to every individual. While many have been successful, most requires complex filtering and processing. Our proposed method is based on a simple estimate of the joint probability of a pair of pixel intensities in predetermined relative positions in the image, also called gray-level co-occurrence matrix (GLCM). First, eye images of different human subjects are obtained. The images are then unwrapped or transformed to the polar space and then segmented using dynamic optimal partitioning in order to isolate the iris features from the rest of the image. Contrast normalization is also applied to reduce the effect of non-uniform illumination over different parts of the image. The GLCM of each iris is calculated and normalized to further minimize the effect of constant shift in gray-level intensities. A new representation of the GLCM are then introduced and used to uniquely describe the GLCM. This new representation is based on off-diagonal peaks that depict "busy" areas of rich details. A rule-based approach incorporates these off-diagonal peaks into a feature vector forming the basis for identification. These features extracted for an individual were found to be highly correlated with a variation of his other features and poorly with iris features of other individuals. This suggests that such simple method may be useful in human iris identification