A Patch-Based Approach for Removing Impulse or Mixed Gaussian-Impulse Noise

Abstract

Abstract. In this paper, we address the problem of the restoration of images which have been affected by impulse noise or by a mixture of Gaussian and impulse noise. We rely on a patch-based approach, which requires careful choices for both the distance between patches and for the statistical estimator of the original patch. Experiments are led in the case of pure impulse noise and in the case of a mixture. The method proves to be particularly powerful, especially for the restoration of textured regions, and compares favorably to recent restoration methods.

PARIGI: a Patch-based Approach to Remove Impulse-Gaussian Noise from Images

“…An estimation of p can naturally be derived from this estimation by computing the ratio between the number of noisy pixels and the image size. As explained in [3], the estimation of p in the implementation of PARIGI is done with the detector ROAD.…”

“…More properties of this distance and some comparisons with other possible robust distances were given in [3]. Notice in particular that when p = 0, then the weighted distance D w is simply the L 2 distance.…”

“…An empirical study of the quadratic risk for the estimation of u was made in [3]. Obviously, the risk of the estimator in an ideal case, where the number of samples of a noisy patch is potentially infinite, decreases as n gets larger.…”

“…The aim of this paper is to precisely describe an implementation of the PARIGI denoising method introduced in [3]. PARIGI stands for Patch-based Approach to Remove Impulse-Gaussian noise from Images.…”

“…This paper focuses on the recent patch-based denoising algorithm called PARIGI [3]. The idea of patch-based approaches is to take advantage of the redundancy of patches (also sometimes called the self-similarity) in images.…”