Face recognition via selective denoising, filter faces and hog features

Face recognition has become a very important topic in recent years. In this paper, we introduce selective denoising with block matching and 3D filtering (BM3D) or video BM3D (VBM3D), compute filter faces, and extract the histogram of oriented gradient (HOG) features from the extracted feature maps. We apply our new method to both illumination invariant face recognition and hyperspectral face recognition. For illumination invariant face recognition, our proposed method in this paper achieves the highest correct classification rate (98.4%) for the Extended Yale Face dataset B and similar results (100%) for the CMU-PIE dataset. For hyperspectral face recognition, our new method achieves perfect classification rate (100%) for both the PolyU-HSFD dataset and the CMU-HSFD dataset.