STUDY ON FAST SUPRE-RESOLUTION RECONSTRUCTION METHOD BASED ON SPARSE-REPRESENTATION AND CLUSTERING

In order to improve the speed of image super-resolution reconstruction,this paper proposes a fast super-resolution reconstruction method based on sparse-representation and clustering.Instead of using a global dictionary.we apply the clustering algorithm to the dictionary,so that similar atoms can be assigned to the same cluster.Then,for each given low-resolution image patch,we can select one atom-cluster adaptively to accelerate the sparse decomposition.Experiment shows that our method is able to achieve desirable super-resolution images with significant computational advantages.