Abstract:Iris image segmentation is very important for an iris recognition system. There are always iris noises as eyelash,eyelid,reflection and pupil in iris images. The paper proposes a virtual method of segmentation. By locating and normalizing iris images with Gabor filter,we can acquire information of image texture in a series of frequencies and orientations. Iris noise regions are determined based on phase congruency by a group of Gabor filters whose kernels are suitable for edge detection. These regions are filled according to the characteristics of iris noise. The experimental results show that the proposed method can segment iris images effectively.