Face Recognition Based on Gabor Multi-channel Weighted Optimization and Sparse Representation

Very recently, the sparse representation theory in pattern recognition arouses widespread concern. In this paper, the sparse representation-based face recognition algorithms are studied. In order to make the representation coefficient vector sparser, a Gabor Sparse Representation Classification (GSRC) algorithm is presented, which uses the Gabor local feature to construct dictionary to enhance the robustness for the external environment changes. GSRC algorithm equally treats all the Gabor features, while in consideration that different Gabor features distinctively contribute to the face recognition task, a Weighted Multi-Channel Gabor Sparse Representation Classification (WMC-GSRC) algorithm is further proposed. By introducing the Gabor multi-channel model, WMC-GSRC algorithm extracts Gabor features in different channels to construct dictionaries and sparse representation classifiers, and obtains the final classification result by performing the weighting fusion of classifiers. Experimental results given in the paper on the ORL, AR and FERET face databases show the feasibility and effectiveness of the proposed methods.