Numerical Modeling for the Recovery of Fine Structures in PDE-based Image Denoising

PDE-based image denoising processes often lead to significant loss of fine structures such as edges and textures. This article is concerned with numerical modeling for PDE-based denoising models that can preserve/recover fine structures in the restored image. The essentially non-dissipative schemes are considered to minimize nonphysical diffusion during the removal of impulse noise. For the removal of mean-zero noises, effective strategies are studied for variable constraint parameters that can recover fine structures and optimal timestep sizes that can reduce the noise components more rapidly than other components. Here the goal is to remove combinations of impulse and Gaussian noises, with no useful fine structures lost in the residual. Various examples are presented to show efficiency and reliability of the numerical modeling techniques in image denoising.