AN adaptive L1–L2 hybrid error model to super-resolution

A hybrid error model with L<inf>1</inf> and L<inf>2</inf> norm minimization criteria is proposed in this paper for image/video super-resolution. A membership function is defined to adaptively control the tradeoff between the L<inf>1</inf> and L<inf>2</inf> norm terms. Therefore, the proposed hybrid model can have the advantages of both L<inf>1</inf> norm minimization (i.e. edge preservation) and L<inf>2</inf> norm minimization (i.e. smoothing noise). In addition, an effective convergence criterion is proposed, which is able to terminate the iterative L<inf>1</inf> and L<inf>2</inf> norm minimization process efficiently. Experimental results on images corrupted with various types of noises demonstrate the robustness of the proposed algorithm and its superiority to representative algorithms.