Image Denoising: Issues and Challenges

Abstract

Extracting noise coefficients from images to make them more informative is of critical importance in plenty of applications. There have been a lot of denoising techniques existing in the literature. However, the fidelity of the denoising algorithm is determined by the quality of the reconstructed image. Minimum artifacts and better preservation of geometrical details such as edges and texture reflect efficient image reconstruction. This paper presents a review of some remarkable trends in the field of denoising. The foremost objective of this study is to provide an understanding of research in the context of detail preserving denoising. In this paper, we also explore how denoising challenges force growth of denoising techniques from spatial domain filters to transform domain, from transform domain to hybrid domain and then to dictionary learning based methods. While all these approaches aim at productive denoising, they possess different assumptions, advantages, and shortcomings. This paper also provides us a deep understanding of which method to be adopted to obtain the most reliable results.