Abstract: Different types of denoising methods are existed in databases. But every method has its own uniqueness. We propose a new method that is adaptive patch based system for image denoising. The approach depends on a pointwise selection of narrow image patches of precise size in the variable neighborhood of each pixel. Our contribution is to engage in each pixel the weighted sum of data points not outside an adaptive neighborhood, in a sense that it balances the efficiency of estimation and the stochastic error, at each contiguous position. In this paper introducing spatial adaptivity, we prolong the work that can be designed as a development of bilateral filtering to image patches. This method is tested by using AWGN and images are taken from databases with different resolutions. The performance of the denoising system is computed in terms of PSNR by taking different noise levels.
Keywords: Image denoising; bilateral filtering; mean-squared-error (MSE) estimator; PSNR; Patch based method.