An adaptive diffusion coefficient selection for image denoising

In the gradient dependent denoising methods based on partial differential equation, the process of denoising is controlled through the gradient operation. Hence, the edges are preserved while texture and fine details (having oscillatory nature, the same as noise) are degraded. This paper proposes an algorithm which adaptively selects diffusion coefficient using the residual local power and the amount of the gradient magnitude. Since texture regions correspond to large values of the local power of the residue, this strategy permits to simultaneously preserve the edges, textures, and fine details. To evaluate the proposed method, a variety of experiments are carried out confirming the performance of the proposed algorithm with respect to peak signal-to-noise ratio, mean structural similarity, universal quality index, visual information fidelity and visual quality. We propose an adaptive selection of diffusion coefficient.It utilizes the residual local power to control the diffusion process.This strategy permits to simultaneously preserve edges, textures, and fine details.Experimental results confirm the performance of the proposed method.