融合局部与全局特征提取的虹膜识别方法

摘要

Traditional iris recognition systems convert iris images into polar coordinates, and then normalized the images to achieve rotation invariance by rotating the feature vector. In order to decrease the complexity of typical iris recognition method, this paper presented a method of iris image identification based on global and local features that extracted from preprocessed iris image without normalizing. Firstly,it applied a bank of non-tensor product wavelet filters to extract the global features of iris. Secondly, it used a SIFT method to extract the local features of the selected regions. Finally, it tested the similarity distances of local and global features with different weights. Experimental results show that the proposed method has the correct recognition rate of 99. 065% when the equal error rate is 0. 935% . Without normalizing the iris images, the proposed approach can obtain very good recognition performance.%传统的虹膜识别系统需要将虹膜图像转换至极坐标系统并进行归一化,通过平移特征向量来达到旋转不变性.为了降低传统虹膜识别方法的复杂性,提出了一种融合局部与全局特征提取的虹膜识别方法,无须对预处理后的虹膜图像进行归一化.该方法首先对分割出的虹膜图像直接采用非张量积小波提取全局特征,接着采用SIFT方法提取选定区域的局部特征,最后对虹膜局部及全局特征采用不同的权值,进行相似性距离测试.结果表明该方法在等错误率为0.935％的情况下,正确识别率达到了99.065％.在不对虹膜图像归一化的情况下,可获得很好的识别性能.