【图像去噪】基于回归的非局部均值算法实现图像去噪附matlab代码

1 简介

It was recently demonstrated in  that the denoising performance of Non-Local Means (NLM)

can be improved at large noise levels by replacing the mean by the robust Euclidean median. Numerical experiments on synthetic and natural images showed that the latter consistently performed better than NLM beyond a certain noise level, and signifificantly so for images with sharp edges. The Euclidean mean and median can be put into a common regression (on the patch space) framework, in which the ℓ2 norm of the residuals is considered in the former, while the ℓ1 norm is considered in the latter. The natural question then is what happens if we consider ℓp (0 < p < 1) regression? We investigate this possibility​ in this paper.

2 部分代码

% demo on Non-Local Patch Regression%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%​%%  Reference:%  K. N. Chaudhury, A. Singer, "Non-Local Patch Regression:%  Robust image denoising in patch space", IEEE International%  Conference on Acoustics, Speech, and Signal Processing, 2013.%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%​clc; clear all; close all force;​% clean imageimg    = double(imread('ckb.jpg'));[m, n] = size(img);​% add noisesigma     =  40;imgNoisy  =  img  +  sigma * randn(m,n);​% NLEM parametersp  = 0.1;S  = 10;P  = 3;h  = 10 * sigma;​% call NELM (or NLEM_kNN)imgDenoised = NLPR(imgNoisy, h, P, S, p);​​% show resultspeak  = max(max(img));PSNR0 = 10 * log10(m * n * peak^2 / sum(sum((imgNoisy - img).^2)) );PSNR1 = 10 * log10(m * n * peak^2 / sum(sum((imgDenoised - img).^2)) );​figure('Units','normalized','Position',[0 0.5 1 0.5]);colormap gray,subplot(1,3,1), imagesc(img),title('Original', 'FontSize', 10), axis('image', 'off');subplot(1,3,2), imagesc(imgNoisy),title([ 'Noisy, ', num2str(PSNR0, '%.2f'), 'dB'] , 'FontSize', 10),axis('image', 'off');subplot(1,3,3), imagesc(imgDenoised),title([ 'NLEM filtered, ', num2str(PSNR1, '%.2f'), 'dB'] , 'FontSize', 10),axis('image', 'off');​

3 仿真结果

4 参考文献

Yingkun Hou, Jun Xu, Mingxia Liu, Guanghai Liu, Li Liu, Fan Zhu, and Ling Shao, "NLH: A Blind Pixel-level Non-local Method for Real-world Image Denoising", IEEE Transactions on Image Processing, vol. 29, pp. 5121-5135, 2020.

博主简介：擅长智能优化算法、神经网络预测、信号处理、元胞自动机、图像处理、路径规划、无人机等多种领域的Matlab仿真，相关matlab代码问题可私信交流。

部分理论引用网络文献，若有侵权联系博主删除。