面向虹膜识别的Gabor特征筛选研究

Study of Gabor feature selection for iris recognition

针对2D-Gabor虹膜特征并不稳定,影响虹膜识别率的问题,提出了一种从多尺度、多方向2D-Gabor小波提取的虹膜特征中,筛选稳定特征应用于虹膜识别的方法。对虹膜图像采用多通道Gabor小波提取虹膜图像特征,然后通过自定义筛选准则从多维特征中筛选出最优特征参数并编码,用Hamming距进行特征匹配识别。基于CASIA虹膜图像库进行实验,结果表明该方法扩大了类内匹配与类间匹配之间的Hamming距,降低了等错率,同时降低了编码的长度,加快了特征匹配速度。

In common, the instability of 2D-Gabor iris features decreases the recognition ratio of the iris. To settle the problem, this paper proposes an approach to select the secure features for iris recognition from multiple-scale 2D-Gabor features. The features of iris are extracted by multi-channel Gabor filtering on the iris images. Then the optimal feature parameters are selected by the screening rules defined and coded. And the feature matching is per- formed by using Hamming distance. Experimental results on the CASIA iris database images show that this ap- proach enlarges Hamming distance between inter-class and intra-class, and at the same time the equal error rate （EER） is decreased. Besides, the length of code is reduced and speed of feature matching is improved.