Nonlocal image denoising algorithm based on image statistic

In this paper, we propose a robust and image denoising method based on image statistic analysis. The image statistic method based on Weibull distribution is applied to image patch content analysis. According to the content analysis, image patches are classified into three types: smooth type, edge type and texture type. And then, different patch similarity measure method and measure window size are applied to denoise images with different types of patches. Based on the results from various different images, our content based NL-means algorithm is shown to have better performance in both PSNR and visual quality compared to the traditional NL-mean algorithm.