A New Iris Segmentation Method Based on Improved Snake Model and Angular Integral Projection

Segmenting iris region is fundamental for iris-based biometric systems. The overall performance of an iris recognition system is highly dependent on accurate iris segmentation. In this paper, a new algorithm for iris segmentation is proposed towards more accurate and efficient segmentation, it detects the precise pupil contour and localizes the limbic boundary. An improved snake model is presented, wherein a new external energy function is designed. This external energy is computed based on the Angular Integral Projection Function (AIPF). First, the AIPF is combined with the improved snake to detect the pupil boundary. For that, pupil boundary points are detected by using the AIPF, and circle fitting is followed to localize the circular pupil boundary giving the initial snake contour. Then, the precise pupil contour is detected by deploying the improved snake. Second, as another contribution of this work, the limbic boundary is localized by combining the AIPF with a technique to extract outlier boundary points based on Mahalanobis distance. Experimental results on CASIA V3.0 iris image database show that the improved snake model is comparable with conventional snake model, and the whole segmentation performance of the proposed algorithm outperforms those of other wellknown existing methods in both terms of accuracy and processing time.