Abstract: The super-resolution algorithm based on sparse representation selects atoms in the dictionary by approximately random style to fit the specified image patch,but the selected atoms show strong structure sparsity in practice, which leads to the complexity of calculation and a great deal of errors,affecting the quality of the reconstructed images.For conquering the problem,a online single image super-resolution algorithm based on group sparse representation was proposed.The proposed algorithm takes advantage of the inputted low-resolution image to construct the group sparse dictionary by introducing the group sparse theory,and then incorporates the group sparse prior and geometric duality prior to design the cost function of the algorithm,which is solved by a proposed iterative optimization method.The experiments demonstrate that the proposed algorithm is superior to the main stream algorithms subjectively and objectively.

Key words: Group sparse representation,Single image super-resolution,Orthogonal matching pursuit,Dictionary lear-ning,Iterative optimization