Anisotropic smart shape-adapted image smoothing without conductance function efficient for impulse noise removal

The major challenge facing any denoising method is to suppress all kinds of noise such as distortions and artifacts, while preserving sharp details and edges in the image. To deal with this challenge, we propose, in this paper, a new smart anisotropic method for impulse noise removal. This method is based on a dynamic filter supervised by a fuzzy controller and unbalanced push-pull tensors. The controller converts the traditional denoising process into a dynamic diffusion over homogeneous zones and reduces progressively the filtering strength in order to stop it over edges and borders. The proposed idea consists in developing a self-space varied filter without any edge stopping function. The filter size varies and auto-adapts locally to noise density and image content with the purpose of preserving details based on a buffer zone called "The anisotropic corridor". In this corridor, the filtering process is forbidden over edges and details. The unbalanced tensor forces are applied to dislocate the filter structure and auto-adapt it to the image content. Simulation results and comparative study are conducted with recent papers in the literature to demonstrate its denoising efficiency. This efficiency is furthermore proved by objective criteria for quality measurements. (C) 2018 Elsevier Inc. All rights reserved.