Face Recognition Based on Wavelet Transform and MPCA

The classic Principal Component Analysis consider the image as a whole, but when the image has expression changes or attitude changes in it, the accuracy rate of the PCA algorithm will decrease a lot. This paper will combine the wavelet transform and an improved MPCA algorithm, Firstly the original image was decomposed by wavelet transformation, take the low frequency information of the face image as the sample, then extract the eigenvectors of the sample using the improved MPCA, nearest neighbor classifier is adopted to sort and distinguish the test samples. Experimental results on ORL database indicate that the performance of this paper's algorithm is superior other methods, such as PCA and MPCA algorithm.