Weighted discriminative sparsity preserving embedding for face recognition

摘要

Sparse representation (SR) based dimension reduction (DR) methods have aroused lots of interests in the field of face recognition. In this paper, we firstly propose a new sparse representation method called weighted elastic net (WEN). Compared to the existing SR methods, WEN is able to explore and use the local structures of data sets sufficiently. Based on WEN, a new supervised sparse representation based DR algorithm called weighted discriminative sparsity preserving embedding (WDSPE) is proposed. In WDSPE, the within-class scatter and between-class scatter of a given data set are constructed by using WEN. Consequently, WDSPE seeks a subspace in which the ratio of the between-class scatter to the within-class scatter is maximized. Moreover, by integrating the global discriminative structures of data sets, we present an extension version of WDSPE. Experiments conducted on three popular face databases (Yale, AR and the extended Yale B) with promising results demonstrate the feasibility and effectiveness of the proposed methods.