基于对称的二维主成分分析及其在人脸识别中的应用

在二维主成分分析算法中引入了对称性思想，提出了基于对称的二维主成分分析算法（STDPCA）。在该算法中，首先把人脸图像分解成奇对称图像和偶对称图像，然后分别在这两类图像中进行二维主成分分析，提取所需要的特征。该算法不仅有效利用了二维主成分分析算法的优点，而且也考虑了人脸对称性的特点，因此在人脸识别中有较高的识别率。在著名人脸库ORL、YALE中的实验证明了该算法的有效性。

Symmetry based two-dimensional principal component analysis and its application to face recognition

This paper presented a Symmetry based Two-Dimensional Principal Component Analysis (STDPCA) on the basis of this idea of symmetry, which was introduced into Two-Dimensional Principal Component Analysis (TDPCA). Firstly, facial image was divided into the even symmetrical image and the odd symmetrical image. Then TDPCA was performed in the even symmetrical image and the odd symmetrical image for feature extraction, respectively. Therefore, STDPCA used effectively not only the advantages of TDPCA, but also the symmetrical properties of facial images. The experimental results on ORL and YALE database show the efficiency of STDPCA.