A non-standard anisotropic diffusion for edge-preserving noise removal

The main objective of this article is to develop a nonstandard partial differential equation-based model for efficient edge-preserving image denoising. Standard total variation (TV)-based models can easily lose fine structures and produce non-physical dissipation during the noise removal process. The principal feature in this article is an introduction of a new coefficient on the non-linear diffusion term of the Euler-Lagrange equation corresponding to the minimization of the energy functional. Combination of a new model with a texture-free residual parametrization enables us to overcome the drawback arising from use of the standard TV-based model. The numerical results indicate the effectiveness and robustness of the new model.