基于小波变换的融合两种2DPCA的人像认证算法  

Algorithm of Face Authentication Based on Wavelet Transform and Combining Two Different 2-Dimensional Participial Component Analyses

作　　者：栾方军[1] 郭红梅[2] 蔺蘭[1] 王永会[1] 

机构地区：[1]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168 [2]辽宁大学轻型产业学院,辽宁沈阳110036

出　　处：《沈阳建筑大学学报（自然科学版）》2010年第5期1001-1005,共5页Journal of Shenyang Jianzhu University：Natural Science

基　　金：辽宁省自然科学基金项目(20072009)

摘　　要：目的为了解决单一2DPCA方法在人像识别中的局限性问题,提出一种在小波变换降维的基础上融合两种2DPCA的人像认证算法.方法采用2维小波变换对人像进行分解,抽取低频信息构成新的特征描述,再分别使用基于总协方差矩阵和类间散布矩阵的2DPCA方法进行初次判决,最后融合2组判决结果得出最终判决,实现基于人像的身份认证.结果在ORL数据库上进行实验,小样本和全部样本情况下最高识别率分别达到99.92%和94.25%,表明了算法的有效性.结论算法优于二维主成分分析方法、单一的基于小波变换的二维主成分分析方法及二维独立元方法,且对表情、姿态及光照变化有一定稳定性.Identity verification based on faces,as a kind of biometric authentications,has become one of the hot spots on research in the field of identity verification.In this paper,an algorithm of face verification which is based on wavelet transform and combining two different 2-Dimensional Principle Component Analyses（2DPCAs）is proposed.At first,face image is decomposed by 2-Dimational Wavelet Transform（2DWT）.Then two different 2DPCAs are used to get the results of face recognition respectively.To the end,the final result is given by fusing them.Experiment on ORL database shows that the algorithm is valid.According to the results,the best recognition rates are 99.92% and 94.25% in the case of small-sample and all-sample,respectively.Algorithm of face verification based on wavelet transform and combining two different 2DPCAs is superior to that of 2DPCA,single wavelet transformation 2DPCA and 2DICA.It also maintains certain stability within changes of expression,pose and illumination.

关 键 词：生物特征认证 人像认证 小波变换 2DPCA 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]