Supervised sparse representation with coefficients´ group constraint

Sparse representation has gained much attention of many researchers recently due to the powerful ability of representing and compressing the original sample. The sparse based classification (SRC) method has been proposed for face recognition and applied to many other fields, which method aims to sparse represent test sample on training set and minimize the reconstruction error. In order for better representation, it is expected that the original sample is represented by samples in same class as much as possible. Base on this assumption, in this paper, a group constraint for coefficients is introduced into the object function of sparse representation to penalize the non-zero coefficients with different classes from the original samples class. The function is solved efficiently by the conventional subgradient method. Experiments on several databases from three fields, such as face recognition, digit recognition and natural image classification, demonstrated the effectiveness of the proposed algorithm.