双向压缩的2DPCA与PCA相结合的人脸识别算法

Face recognition combined bidirectional 2DPCA with PCA

主成分分析(PCA)直接用于人脸识别时,需将图像矩阵转换成向量,导致求解高阶矩阵计算量大。二维主成分分析(2DPCA)的实质是对图像矩阵按行进行图像压缩抽取特征,消除了图像列的相关性,但特征数量仍然较大,影响分类速度。针对这一问题,提出了采用双向压缩的二维主成分分析消除图像行间和列间的相关性,再结合PCA进一步减少特征数量,改进人脸识别算法,该算法用于ORL人脸库上得到了较高的识别率和较快的识别速度。

The method of Principal Component Analysis （PCA） needs to translate matrix into vectors directly used in face recognition, it results in large computation calculating high - rank matrix. The essence of traditional Two-Dimensional Principal Component Analysis （2DPCA） is to extract features of image matrix using PCA in each row, it eliminates relativity between columns, but the number of features is still large, it affects the speed of classification. To figure out this problem, the author adopted bidirectional 2DPCA to eliminate relativity between columns and between rows, then used PCA to reduce the number of features again, using this way on the ORL human face libraries, it gets upper recognition rate and faster speed.