A Novel Compound Approach for Iris Segmentation

With a growing emphasis on human identification, iris recognition as a biometric identification has recently received increasing attention. The first step in iris recognition is segmentation. In this paper a new segmentation approach is offered which does not use any information of color or texture as the segmentation cues. Instead, we use random sample consensus method to fit an ellipse or a circle to the edge information of iris boundary. The presented approach is robust against the iris texture variations and other trouble makers like eyelid and specularity effect in pupil area. The extracted curves in this method are more conformable with iris boundaries than the curves obtained by other conventional methods.