基于ADMM的TV正则化稀疏图像重建，Matlab代码如何实现？

本文共计469个文字，预计阅读时间需要2分钟。

1. 简介

2.编辑

3.部分

4.代码

5.This simple demo examines if TVAL3 works normally. Please try more demos in the Demos directory, which would show users what TVAL3 is capable of.

6.I: 64x64 phantom (real, two-dimensional)

7.A: ran

1 简介

​

编辑

2 部分代码

% This simple demo examines if TVAL3 works normally. Please try more demos
% in the "Demos" directory, which would show users what TVAL3 is capable of.
%
% I: 64x64 phantom (real, two-dimentional)
% A: random matrix without normality and orthogonality (real)
% f: observation with/without noise (real)
%
clear; close all;
path(path,genpath(pwd));
fullscreen = get(0,'ScreenSize');
% problem size
n = 64;
ratio = .21;
p = n; q = n; % p x q is the size of image
m = round(ratio*n^2);
% sensing matrix
A = rand(m,p*q)-.5;
% original image
I = phantom(n);
nrmI = norm(I,'fro');
figure('Name','TVAL3','Position',...
[fullscreen(1) fullscreen(2) fullscreen(3) fullscreen(4)]);
subplot(121); imshow(I,[]);
title('Original phantom','fontsize',18); drawnow;
% observation
f = A*I(:);
favg = mean(abs(f));
% add noise
f = f + .00*favg*randn(m,1);
%% Run TVAL3
clear opts
opts.mu = 2^8;
opts.beta = 2^5;
opts.tol = 1E-3;
opts.maxit = 300;
opts.TVnorm = 1;
opts.nonneg = false;
opts.disp = false;
opts.TVL2 = false;
t = cputime;
[U, out] = TVAL3(A,f,p,q,opts);
t = cputime - t;
subplot(122);
imshow(U,[]);
title('Recovered by TVAL3','fontsize',18);
xlabel(sprintf(' %2d%% measurements \n Rel-Err: %4.2f%%, CPU: %4.2fs ',ratio*100,norm(U-I,'fro')/nrmI*100,t),'fontsize',16);

3 仿真结果

​

编辑

4 参考文献

[1]徐敏达, 李志华. 基于L1与TV正则化的改进图像重建算法[J]. Computer Science, 2018, 45(012):210-216.

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。

​

编辑