基于纹理分布特征的虹膜识别算法.pdfVIP

第 31 卷 第 2 期 仪 器 仪 表 学 报 Vol.31 No. 2 2010 年 2 月 Chinese Journal of Scientific Instrument Feb. 2010 基于纹理分布特征的虹膜识别算法* 苑玮琦，赵彦明，张志佳 （沈阳工业大学视觉检测技术研究所 沈阳 110178） 摘 要：二维小波变换方向性差，不能从方向和频率同时描述虹膜纹理。基于此，分析了归一化虹膜纹理的分布特点，提出 了基于纹理分布特征的虹膜识别算法。对原始人眼图像进行虹膜定位等预处理操作，得到归一化的虹膜纹理图像；对归一化 虹膜图像进行了二维小波多尺度变换，结合虹膜纹理分布特点，选择小波分解水平通道；依据点能量贡献度，消除伪特征点 并进行了点能量编码；计算了不同虹膜编码间海明距离，以其为依据进行分类。在给定距离阈值为 0.25 的前提下，可以达 到 99.91% 的正确识别率。实验证明该算法是有效、可行的，并具有较高的识别率，识别速度也很快。 关键词：纹理分布；虹膜识别；小波变换；多尺度；虹膜 中图分类号：TP391.41 文献标识码：A 国家标准学科分类代码：460.4030 Iris recognition algorithm based on texture distribution feature Yuan Weiqi, Zhao Yanming, Zhang Zhijia (Computer Vision Group, Shenyang University of Technology, Shenyang 110178, China) Abstract ：2D wavelet transformation is weak in direction and can’t describe iris texture in direction and frequency at the same time. Thus, this paper analyzes the features of iris texture distribution and proposes an iris recognition algorithm based on texture distribution feature. Firstly, the algorithm carries out preprocessing of the original eye image such as iris localization and gets the normalized iris features image; secondly, 2D wavelet transformation is performed on the normalized iris images. Through combining the features of iris texture distribution, the horizon- tal channel of wavelet decomposition is selected; and according to the energy weights, the pseudo-feature points are deleted and point energy coding is carried out. Thirdly, the Hamming distances between different iris codes are calculated and then classified based on the distances. When the distance threshold equals to 0.25, the accuracy is 99. 91%. Related experiments show that the iris recognition algorithm based on wavelet mult