计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (24): 181-183.DOI: 10.3778/j.issn.1002-8331.2008.24.055

• 图形、图像、模式识别 • 上一篇    下一篇

基于多通道小波滤波的虹膜识别新算法

王凤随,冯 桂   

  1. 华侨大学 信息科学与工程学院,福建 泉州 362021
  • 收稿日期:2007-10-22 修回日期:2008-01-02 出版日期:2008-08-21 发布日期:2008-08-21
  • 通讯作者: 王凤随

Novel iris recognition algorithm based on multichannel wavelet filter

WANG Feng-sui,FENG Gui   

  1. College of Information Science and Engineering,Huaqiao University,Quanzhou,Fujian 362021,China
  • Received:2007-10-22 Revised:2008-01-02 Online:2008-08-21 Published:2008-08-21
  • Contact: WANG Feng-sui

摘要: 为克服小波变换过零检测虹膜识别算法对纹理灰度变化敏感的缺点,利用小波变换多通道滤波的特性,提出了一种新的基于Daubechies-4小波的虹膜识别新算法。根据虹膜纹理分布的特征,将虹膜分成10个分析带,对每个分析带采用一个合适尺度的小波滤波,小波变换各个通道的小波系数的均值及标准差作为虹膜的特征值,最后得到虹膜的128位特征编码。特征匹配采用的是加权欧式距离的分类器方法。实验结果表明算法是有效的,取得了高识别率;同时,提出的算法对虹膜图像的尺度、旋转、平移等的变化具有不变性。

关键词: 虹膜识别, 多通道, 小波变换, 纹理分析, 特征提取

Abstract: Iris recognition algorithms based on wavelet transform zero-crossing are sensitive to the grey variation of the iris image texture.A new iris recognition algorithm based on Daubechies-4 wavelet is presented using the multiple channel characteristic of wavelets.After dividing iris into ten strips according to its texture distributing,each strip adopts an appropriate scale wavelet used to filter each strip.The feature value is denoted by the mean and standard deviation of every channel’s coefficients of wavelet transform,and 128 feature values can be ascertained for every iris.Then the weighed Euclid distance is used to match two iris codes.Tests show that presented method which is translation,rotation,and scaling invariant is effective and achieves a high recognition rate.

Key words: iris recognition, multiple channels, wavelet transform, texture analysis, feature extraction