Local Occluded Face Recognition Based On 2d-Dwt And Sparse Representation

Because most of the current face recognition systems do not consider the possible occlusion in the real environment, the traditional face recognition algorithm has poor recognition results. This paper proposes a face recognition method based on sparse representation of two-dimensional discrete wavelet transform in feature subspace for the presence of facial occlusion. Through two-dimensional discrete wavelet decomposition of the training samples, the high-frequency signals are filtered and the low-frequency signals are retained. The low-frequency signals are constructed by PCA, and then the test samples are sparse decomposed on the occlusion dictionary, and finally classified. Compared with other traditional face recognition methods, the proposed algorithm has better recognition results.