Iris segmentation is considered as the most difficult and fundamental step in an iris recognition system. While iris boundaries are largely approximated by two circles or ellipses, other methods define more accurately the iris resulting in better recognition results. In this paper we propose an iris segmentation method using Hough transform and active contour to detect a circular approximation of the outer iris boundary and to accurately segment the inner boundary in its real shape motivated by the fact that richer iris textures are closer to the pupil than to the sclera. Normalization, encoding and matching are implemented according to Daugmans method. The method, tested on CASIA-V3 iris images database is compared to Daugmans iris recognition system. Recognition performance is measured in terms of decidability, accuracy at the equal error rate and ROC curves. Improved recognition performance is obtained using our segmentation model proposing its use for better iris recognition system.