An improvement on SURE based non-local means

Non-local means algorithm is a powerful image denoising method which Because of its naive realization has attracted a lot of attention from image processing community. Two of the major problems with this method are parameter adjustment and residual noise near sharp edges. In this paper we use Stein´s Unbiased Risk Estimate (SURE) and local variance to overcome these difficulties. As experimental results confirm, proposed method presents high quality results from both objective and subjective point of views.