Local and Non-Local Feature-Based Kernel Nonnegative Matrix Factorization Method for Face Recognition
Wen-Sheng Chen, Qian Wang, Binbin Pan, Yugao Li · 2016
Based on the kernel method and graph theory, this paper proposes a novel Kernel Non-negative Matrix Factorization with Local and Non-local feature (LN-KNMF) approach for face recognition. We establish the objective function in kernel space which incorporates two scatter quantities, namely local scatter and non-local scatter. They are determined by the local adjacent graph matrix and non-local adjacent graph matrix respectively. The update rules of the proposed LN-KNMF method are derived using polynomial kernel function and gradient descent method. Subsequently, we theoretically prove the convergence of the proposed algorithm by means of a constructed auxiliary function. Experimental results on the ORL face database demonstrate the superior performance of our approach against some NMF-based methods.