Impulsive Noise Removal by Nonlocal Fuzzy Diffusion

Standard image deconvolution methods perform well with the Gaussian noise model based on the piecewise smooth assumption. But the performance on impulsive noise is poor because the noised image is far away from assumption of piecewise smooth. The challenge is how to handle the discontinuities and at the same time remove the impulsive noise. Median-based approaches are inadequate, because at high noise levels they induce nonlinear distortion that hampers the denoising process. Here an improved nonlocal diffusion method based on fuzzy metric is proposed. The experiments results indicate that the new method is suitable for impulsive noise.