Image denoising using collaborative patch-based and local methods

In this paper local and non-local denoising methods are jointly employed in order to improve the visual quality of the final denoised image. Based on the evidence that the output images of non local denoising methods are not pointwise better everywhere than the outputs images of local methods and than the noisy image itself, the cascade of two improvement steps is applied to the output image of a non local denoising method. The first step aims at correcting the output image by recovering the lost information directly from the noisy one. The second step aims at recovering those good estimations provided by a local regularization method. A pointwise weighted average between the involved image pair is used at each step. The weights are estimated from the noisy data using adaptive and automatic procedures. Experimental results show that the proposed approach allows us to greatly improve the results of patch based non local denoising in terms of both peak signal to noise ratio (PSNR) and structural similarity index (SSIM).