Single image super-resolution combined with CNN and sparse representation

Usually, the high-resolution images reconstructed by the traditional single image super-resolution algorithm often have the problems of insufficient high-frequency information and blurred edges. In order to improve the quality of reconstructed images, we propose a single image super-resolution reconstruction method based on a convolutional neural network and sparse representation. First of all, the dictionary learning method of the convolutional neural network trains the input low-resolution image to obtain LR image blocks. Then, the Learned Iterative Shrinkage and Thresholding Algorithm(LISTA) network trains LR image blocks to generate the corresponding sparse coefficients. Finally, in the reconstruction step, corresponding high-resolution image patches are obtained by Multiplying low-resolution image patches with optimal sparse coefficients. Then corresponding high-resolution image patches are combined to get the final HR image. Experimental results show that the proposed method can effectively combine the two advanced image super-resolution methods to form an efficient and flexible single image super-resolution network.