Impulsive noise removal from images using sparse representation and optimization methods

In this paper, we propose a new method for impulsive noise removal from images. It uses the sparsity of natural images when they are expanded by mean of a good learned dictionary. The zeros in sparse domain give us an idea to reconstruct the pixels that are corrupted by random-value impulse noises. This idea comes from this reality that noisy image in sparse domain of original image will not have a sparse representation as much as original image sparsity. In this method we assume that the proper dictionary to achieve image in sparse domain is available.