基于直方图统计特征的虹膜粗分类  

Coarse Iris Classification Using Histogram Statistics Features

作　　者：唐荣年[1] 翁绍捷[1] 王勇[2] 

机构地区：[1]海南大学机电工程学院,海南儋州571737 [2]西安电子科技大学电子工程学院,西安710126

出　　处：《光电工程》2011年第1期146-150,共5页Opto-Electronic Engineering

基　　金：教育部高等学校博士学科点专项科研基金资助项目(200807011007);海南省教育厅高等学校科学研究资助性项目(Hjkj2010-04);海南大学2009年度科研项目(hd09xm82)

摘　　要：为改善大规模虹膜数据库中虹膜匹配算法的性能,本文提出了一种基于灰度统计特征的虹膜类型识别算法。该算法首先将虹膜纹理区域划分为8个子块,然后利用直方图比率提取虹膜类型的直方图统计特征。最后,根据对应子块特征的相似度将虹膜图像识别为五种类型。所提算法在CASIA虹膜库中挑选的500幅虹膜图像样本上进行了测试。仿真实验结果表明,所提算法对虹膜粗分类的准确率达到了98.2%。相对原有的Daugman虹膜匹配策略而言,所提算法结合Daugman算法的虹膜匹配策略不仅可以降低虹膜匹配的等错误率,而且使得匹配所用的时间下降了29.4%。Aimed to improving the performance of iris matching, a new method for the coarse classification of iris images using Histogram statistics features is proposed. First, the iris image is segmented into eight blocks. Then, we calculate the Histogram ratio of these image blocks and take them as the class feature of iris image type. Finally, the iris images are classified into five categories in accordance with the similarity of blocks. The proposed method has been tested and evaluated on 500 iris cases from CASIA iris database. The experiment results shows that the proposed method classifies iris types with an accuracy of 98.2%. Compared with the initial Daugman iris matching method, a modified matching method, the proposed method with Daugman method, has lower equal error rate and decreases 29.4% matching time.

关 键 词：统计特征 虹膜匹配 虹膜粗分类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]