Restoration of Images Corrupted by Impulse Noise and Mixed Gaussian Impulse Noise using Blind Inpainting

摘要

This article studies the problem of image restoration of observed imagescorrupted by impulse noise and mixed Gaussian impulse noise. Since the pixelsdamaged by impulse noise contain no information about the true image, how tofind this set correctly is a very important problem. We propose two methodsbased on blind inpainting and $\ell_0$ minimization that can simultaneouslyfind the damaged pixels and restore the image. By iteratively restoring theimage and updating the set of damaged pixels, these methods have betterperformance than other methods, as shown in the experiments. In addition, weprovide convergence analysis for these methods, these algorithms will convergeto coordinatewise minimum points. In addition, they will converge to localminimum points (or with probability one) with some modifications in thealgorithms.