Metaface Block Sparse Bayesian Learning for Face Recognition

Face recognition is an active and challenging task in pattern recognition and computer vision application. Sparse representation based classification has been verified to be powerful for face recognition. This paper proposes the metaface block sparse bayesian learning (MBSBL) based on the framework of sparse representation. The MBSBL combines the metaface learning and block sparse bayesian learning together to treat the face recognition problem. After obtaining a trained optimal dictionary by the method metaface learning, the representation coefficient is solved through block sparse bayesian learning with the trained dictionary. Experimental results demonstrate that the MBSBL can combine the advantages of metaface learning and block sparse bayesian learning well to achieve better recognition rates.