基于改进的2DPCA人脸识别方法研究

Research on Human Face Recognition Based on Improved 2DPCA Method

传统的主成分分析(PCA)方法在图像识别时需将图像矩阵转化成向量,造成图像向量的维数偏高,使得整个特征提取过程的计算量较大;在PCA的基础上,有人提出了二维主成分分析(2DPCA)的方法,但其本质是对图像矩阵按行进行特征提取,虽然消除了图像列的相关性,但是仍然忽视了行的相关性;因此,在此考虑一种改进的方法能同时消除图像行、列的相关性,并通过实验得到了比2DPCA更高效的识别率。

Traditional principal component analysis （PCA） needs to transform image matrix into vector during image identification, resulting in higher dimension of image vector and making larger amount of computation in the process of the entire feature extraction. On the basis of PCA, two-dimensional principal component analysis （2DPCA） method was pointed out, but its essence is to make feature extraction of image matrix by row, although the relevance of its image column is eliminated, the relevance of a row is ignored. Thus, a kind of improved method for simultaneously eliminating the relevance of row and column of the image is considered, the higher efficient recognition rate than 2DPCA is obtained by experiments.