Face Recognition Using Topology Preserving Nonnegative Matrix Factorization

In this paper, a novel Topology Preserving Nonnegative Matrix Factorization (TPNMF) method is proposed for face recognition. The TPNMF is based on minimizing the constraint gradient distance, compared with L2 distance, the gradient distance is able to reveal latent manifold structure of face patterns. Compared with PCA, LDA and original NMF which search only the Euclidean structure of face space, TPNMF finds an embedding that preserves local topology information, such as edges and texture. In the way, the proposed TPNMF method is robust for variable in lighting and facial expression. Experimental results show that the proposed TPNMF approach provides a better representation of face patterns and achieves higher recognition rates in face recognition.