Non-Local Means Image Denoising with Local Geometric Structures Matching Strategy

Efficient and valid image denoising method is still a challenge, and Non-local means is a rising and effective image denoising method We propose a NL-Means image denoising algorithm based on self-similarity of local geometric structures in the image. As local contrast is an important geometric property of digital image, our algorithm can find patches which are similar not only in grey value but also in local contrast. We demonstrate the application of this approach to enhance the edge and corner structures by simulation experiments, and the result shows that our algorithm increases the PSNR effectively.