Heterogeneous face recognition based on super resolution reconstruction by adaptive multi-dictionary learning

Abstract

The heterogeneous face recognition algorithm is based on super resolution reconstruction and two-dimensional marginal fisher analysis. In this paper, a super resolution reconstruction algorithm by adaptive multi-dictionary learning is adopted. Compared with the traditional global dictionary learning, this algorithm spends less time on dictionary training and image reconstruction to a great extent. Firstly, a sketch is transformed to a photo by eigenface algorithm. Secondly, super resolution reconstruction by improved adaptive multi-dictionary learning is used to reconstruct the synthesized photo, which is able to enhance the quality of synthesized photo effectively. Finally, the synthesized photo is recognized by two-dimensional marginal fisher analysis. We demonstrate these ideas in practice and show how they lead to faster operation speed and ideal recognition rate.