二维主元分析在人脸识别中的应用研究

结合二维主元分析（two-dimensional principal component analysis，2DPCA）的特点，将2DPCA算法用于人脸识别。它与主元分析（principal component analysis，PCA）的不同之处在于，2DPCA算法以图像矩阵为分析对象；而PCA算法以图像的一维向量为分析对象。2DPCA算法是直接利用原始图像矩阵构造图像的协方差矩阵。而PCA算法需对原始图像矩阵先降维、再将降维矩阵转换成列向量，然后构造图像的协方差矩阵。为了测试和评估2DPCA算法的性能，在ORL（olivetti research laboratory）与Yale人脸数据库上进行了实验，结果表明，2DPCA算法用于人脸识别的正确识别率高于PCA算法。同时，也显示了2DPCA算法在特征提取方面比PCA算法更有效。

Application study for 2DPCA in face recognition

Combined with the characteristics of two-dimensional principal component analysis(2DPCA),2DPCA algorithm is applied in face recognition.Unlike principal component analysis(PCA),2DPCA is based on two-dimensional image matrix;whereas PCA is based on one-dimensional vector.By way of 2DPCA algorithm,the original image matrix is directly utilized to construct covariance matrix.In PCA,the dimension of the original image matrix need be reduced firstly;then the reduced-dimensional matrix is converted into vector;afterwards the covariance matrix of the image is obtained.In order to test and estimate the performance of 2DPCA algorithm,experiments on ORL(olivetti research laboratory) and Yale face database are done.Results demonstrate that correct recognition rate of 2DPCA in face recognition is higher than that of PCA;and that 2DPCA is more valid than PCA in feature extraction.