Abstract:A new algorithm named wavelet transform weighted modular PCA is proposed for face recognition. Firstly, the training images and the testing image are preprocessed with wavelet transform and the LL band and the LH/HL average band are divided into sub-images with the same size. Secondly, the prospective classify contribution of each sub-model of the two bands are computed. Thirdly, each sub-image of the two bands of the testing image is projected to its corresponding subspace and the confidence values with each image are obtained. Finally, the two confidence values with each image are added with a weight and the total confidence value is obtained to classify the testing image. Experimental results show that the recognition rate of the proposed algorithm is about 4%-6% superior to traditional methods.

Face Recognition Based on Wavelet Transform Weighted Modular PCA