An Improved Non-local Filter for Image Denoising Using Steering Kernel Regression

Abstract

An improved Non-local Means (NLM) method for image denoising is proposed in this paper. It employs the Rotated Block Matching (RBM) scheme and the adaptive kernel to achieve the robust similarity measure between image patches. The structure of the image obtained by Steering Kernel Regression (SKR) is employed for the RBM and the weighted distance computation. The RBM process can find more similar patches via dominant orientation alignment. Instead of using Gaussian kernel, the SKR kernel (weights) can ensure those of patches with similar structure to get smaller similarity distance values in computing the weighted distance. Finally, the filtering parameter is optimized to obtain better denoisng performance. The proposed method can robustly measure similarity between image patches even if they appear in the rotated instances. Hence, more candidates can be found for the weighted average and yield improved results.