An Effective Iris Recognition System Based on Local Multi-resolution Feature Extraction
Guang Huo · Journal of Information and Computational Science · 2014
Compared with previous designs, the one proposed in this paper has three novelties. Firstly, iris texture is usually occluded by eyelid, eyelash, et al. To eliminate these interference, a novel feature descriptor is proposed – only non-occluded region is extracted by forming overlapping sectors in variable sizes. Secondly, according to the distribution of the iris texture of each sector, appropriate scale is selected for normalization so as to solve the problem of 2D-Gabor filter’s local optimum in a single resolution. Thirdly, because of the normalization of different scales, the overlapping region between the sectors are shown in multi-resolution, so that a better extraction of the scale is obtained which is not easy to be found in a single one. Comparison is made between our algorithm and that of Daugman and Yao in different iris databases. The experimental results show that the proposed system can be more capable of efficient iris feature extraction with a higher accuracy and a better robustness.