NLM Denoising Method with Adaptive Center Pixel Weights

The non-local means (NLM) is an effective and popular denoising method that adjusts each pixel value with a weighted average of all pixels in the entire image. However, the center pixel weights (CPW) in the traditional NLM method and its variances are unitary, and thus the importance of the center pixel is overestimated for the noise point, which cannot effectively remove noises. To address this problem, we propose an adaptive CPW for NLM method. In order to effectively distinguish edges from regions and noises, a new edge indicator is constructed to identify the local characteristic of each pixel. Based on the proposed edge indicator, we construct an adaptive CPW that can be tuned adaptively according to each pixel's local feature. Experimental results show that the propose d method is superior to the state-of-the-art methods in both the edge preservation and noise suppression.