基于二维广义主成分分析的人脸识别

Face recognition based on two-dimensional image principal component analysis

针对人脸识别中的特征提取问题,提出了一种基于二维广义主成分分析(2DIMPCA)的人脸识别方法.有别于传统的人脸识别算法需要将二维人脸图像矩阵压缩成一维向量,该方法直接采用二维图像矩阵来构建方差矩阵,通过在水平和垂直2个方向上顺序执行2次广义主成分分析(IMPCA)运算,消除了人脸图像行和列的相关性,大大压缩了特征的维数.选用二维最小近邻分类法进行分类,计算识别率.在AT&T人脸库以及Yale人脸库上的测试结果表明,与主成分分析(PCA)和IMPCA相比,该方法具有更高的识别率和更快的识别速度.

A new image feature extraction and recognition method based on two-dimensional image principal component analysis （2DIMPCA） was proposed. The method constructed covariance matrices using original image matrices directly and provided a sequentially optimal image compression mechanism. 2DIMPCA performed image principal component analysis （IMPCA） twice, one was in horizontal direction and the other in vertical direction. Also, 2DIMPCA suggested a feature selection strategy to select the most discriminative features. The method was tested and evaluated using the AT＆T face database and the Yale face database. K-nearest neighborhood （KNN） algorithm was used to construct classifiers. Experimental results show that 2DIMPCA is more powerful and efficient than principal component analysis （PCA） and IMPCA for face feature extraction and recognition.