Structure Constrained Discriminative Non-negative Matrix Factorization for Feature Extraction

摘要

In this paper, we propose a novel algorithm called Structure Constrained Discriminative Non-negative Matrix Factorization (SCDNMF) for feature extraction. In our proposed algorithm, a pixel dispersion penalty (PDP) constraint is employed to preserve spatial locality structured information of the basis obtained by NMF. At the same time, in order to improve the classification performance, intra-class graph and inter-class graph are also constructed to exploit discriminative information as well as geometric structure of the high-dimensional data. Therefore, the low-dimensional features obtained by our algorithm are structured sparse and discriminative. Moreover, an iterative updating optimization scheme is developed to solve the objective function of the proposed SCDNMF. The proposed method is applied to the problem of image recognition using the well-known ORL, Yale and COIL20 databases. The experimental results demonstrate that the performance of our proposed SCDNMF outperforms the state-of-the-art methods.