Images Denoising with Feature Extraction for Patch Matching in Block Matching and 3D Filtering

摘要

In this paper, we propose a new method for grey scale image denoising. Our method takes advantage of the fact that the mean of the Gaussian white noise is zero. For every patch in the noisy image, we use a line to divide the image into two regions with equal area, and then take the mean of one of the two regions. We select lines with different slopes in order to extract a number of features. We use these extracted features to match the patches in the noisy image. All other steps in our method are the same as those in the standard BM3D. Our experimental results show that our new method outperforms the standard BM3D for σn>120, and they are identical, otherwise.