PROJECT TITLE :

Affine Non-Local Means Image Denoising - 2017

ABSTRACT:

This paper presents an extension of the Non-Local Means denoising method, that effectively exploits the affine invariant self-similarities present in the pictures of real scenes. Our technique provides a higher image denoising result by grounding on the fact that in several occasions similar patches exist within the image but have undergone a metamorphosis. The proposal uses an affine invariant patch similarity measure that performs an appropriate patch comparison by automatically and intrinsically adapting the size and shape of the patches. Hence, a lot of similar patches are found and appropriately used. We have a tendency to show that this image denoising methodology achieves top-tier performance in terms of PSNR, outperforming consistently the results of the regular Non-Native Means, and that it provides state-of-the-art qualitative results.