基于伪模块2D PCA的人脸识别方法

在模块2D PCA方法的基础上提出了伪模块2D PCA的人脸识别方法.该方法不仅保留了模块2D PCA方法在特征抽取之前无需将图像矩阵转化为图像向量、能快速降低鉴别特征的维数、可以完全避免使用矩阵的奇异值分解等优点,而且在降维的同时尽可能保持了原样本的变化信息,使得降维后的同类数据样本尽可能保持相似.在ORL人脸数据库上的实验结果表明,伪模块2D PCA在识别性能上优于模块2D PCA.

Face recognition method based on pseudo modular 2D PCA

A novel face recognition method,pseudo modular 2D PCA,was proposed based on the modular 2D PCA.The pseudo modular 2D PCA shares the following advantages with the modular 2D PCA: There is no need to transform the image matrices into the image vectors before the feature extraction;The dimension reduction of identification features can be done quickly;The singular value decomposition of the matrices can be absolutely avoided.Meanwhile,the propose method can preserve the variation information of the original samples during the process of the dimension reduction,and it can keep the similarity among the samples of the same class after the process of the dimension reduction.The experimental results of ORL face image database demonstrate that the recognition performance of the pseudo modular 2D PCA is better than that of the modular 2D PCA.