Iris Recognition with Multi-Scale Edge-Type Matching

In this paper, we propose a novel descriptor which characterizes an iris pattern with multi-scale step/ridge edge-type (ET) maps. The ET maps are determined with the derivative of Gaussian (DoG) and the Laplacian of Gaussian (LoG) filters. There are two major advantages of our approach. First, both the feature extraction and the pattern classification are simple and efficient. The iris pattern classification is accomplished by ET matching. The matching of each ET flag can be regarded as a weak classifier and the final decision is based on the vote of each weak classifier. Second, the number of free filter parameters is only three, and hence they can be easily determined. Furthermore, we propose a method for designing the parameters of the filters with the genetic algorithm. The experimental results show that our approach can achieve a recognition rate of 99.98% which is comparable to that of the Gabor filter approach