基于图像矩阵变换的PCA及人脸自动识别  

Principal Component Analysis(PCA) Based on Changing Image Matrices and Face Recognition

作　　者：刘永俊[1] 谈云飞[1] 朱晓宇[1] 常晋义[1] 

机构地区：[1]常熟理工学院计算机学院软件工程系,江苏常熟215500

出　　处：《常熟理工学院学报》2009年第2期101-105,共5页Journal of Changshu Institute of Technology

基　　金：常熟理工学院青年基金(KY2008111)资助项目.

摘　　要：在图像主分量分析的基础上,提出了一种基于图像矩阵变换的主分量分析方法.该方法首先对图像矩阵进行适当的变换,用得到的新的图像矩阵构造图像总体散布矩阵后,再运用图像投影主分量分析进行特征抽取.该方法在ORL标准人脸库上的试验结果表明,经过适当的变换后抽取的鉴别特征在识别性能和速率上均优于单纯的图像主分量分析方法.另外,在AR人脸库上的试验结果也表明该方法对光照变化具有较强的鲁棒性.A new image feature extraction method based on changing image matrices and image principal component analysis （CIMPCA） is proposed in this paper. It has an applicable change of the primitive image matrices, so that the dimension of the image total scatter matrix is further minified. Then, image principal component analysis （IMPCA）is used to do image feature extraction. Finally, the experimental results testing on ORL database indicate that the proposed method is more effective than just using IMPCA simply. Its consumed time for feature extraction is also has a further decreased. Moreover, the experiments on AR database also demonstrate that it is robust in uncontrolled lighting condition.

关 键 词：图像矩阵变换 图像主分量分析 主分量分析 特征抽取 人脸识别 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]