Low-rank representation based robust face recognition by two-dimensional whitening reconstruction

摘要

Our work brings a novel viewpoint and way to combine two-dimensional whitening reconstruction (TWR) with low-rank representation (LRR) for better recognition of faces. Numerous experiments on multiple face databases doubtlessly shows that TWR is beneficial to better recognition of faces. The main reason is that TWR discard several vectors corresponding to the smallest eigenvalues to get optimal or sub-optimal results when constructing whitened faces, and the small eigenvalues might be corresponding to the noise, which usually have severe negative effects on the accuracy of face recognition. Moreover, TWR could transfer each face image to be close to a Gaussian signal, which is more suitable for PCA processing.