资源简介
随着压缩感知技术的发展,基于压缩感知的图像融合技术研究受到越来越多的重视。针对图像傅里叶变换系数特,,提出了一种双星采样模式下基于高低频重要性度量的压缩传感域图像融合算法。该算法首先通过双星采样模式获得测量值,然后计算高低频区域对应的重要性度量作为融合算子,并对测量进行加权融合,最后通过求解最小全变分优化问题重构融合图像。实验结果表明,该算法优于其他基于傅里叶的方案。
代码片段和文件信息
% GenDataFig14: Generate images for figure 14:
% Multiscale CS reconstruction of a Mondrian painting with BPFDRFAR.
I0 = imread(‘mondrian.tif‘);
I0 = double(I0);
% Set finest coarsest scales
j1 = 9;
j0 = 5;
qmf = MakeONFilter(‘Symmlet‘8);
Narr = [0.8 0.6 0.4 0.2];
S = 30;
% Do Multi-CS scheme:
% Sample 4^j0 resume coefficients (coarse-scale
% coeffs) at scale 2^(-j0) x 2^(-j0)
alpha0 = FTWT2_PO(I0 j0 qmf);
alpha_BP = zeros(size(alpha0));
alpha_BP(1:2^j01:2^j0) = alpha0(1:2^j01:2^j0);
alpha_FDR = alpha_BP;
alpha_FAR = alpha_BP;
N_CS = 4^j0;
t_BP = 0;
t_FDR = 0;
t_FAR = 0;
thetas = cell(j1-j01);
alphas_BP = cell(j1-j01);
alphas_FDR = cell(j1-j01);
alphas_FAR = cell(j1-j01);
% For each scale apply CS scheme
for jj = (j0):(j1-1)
% Construct the vector theta of detail wavelet
% coeffs on scale jj
theta1 = alpha0((2^jj+1):2^(jj+1)1:2^jj);
theta2 = alpha0(1:2^(jj+1)(2^jj+1):2^(jj+1));
n1 = prod(size(theta1));
n2 = prod(size(theta2));
theta = [theta1(:); theta2(:)];
thetas(jj-j0+1) = {theta};
Mdetail = 4^(jj+1) - 4^jj;
Ndetail = floor(Mdetail .* Narr(jj-j0+1));
N_CS = N_CS + Ndetail;
% Sample Ndetail compressed samples using the CS operator
y = FastCSOperator(1NdetailMdetailtheta1:MdetailMdetail);
% Solve the CS problem with BP
tic
alpha = SolveBP(‘FastCSOperator‘ y Mdetail);
t_BP = t_BP + toc;
alpha_BP((2^jj+1):2^(jj+1)1:2^jj) = reshape(alpha(1:n1) 2^(jj+1)-2^jj 2^jj);
alpha_BP(1:2^(jj+1)(2^jj+1):2^(jj+1)) = reshape(alpha(n1+1:n1+n2) 2^(jj+1) 2^(jj+1)-2^jj);
alphas_BP(jj-j0+1) = {alpha};
% Solve the CS problem with FDR thresholding
q = 0.9;
tic
[alpha iters] = SolveStOMP(‘FastCSOperator‘ y Mdetail ‘FDR‘ q S 1);
t_FDR = t_FDR + toc;
alpha_FDR((2^jj+1):2^(jj+1)1:2^jj) = reshape(alpha(1:n1) 2^(jj+1)-2^jj 2^jj);
alpha_FDR(1:2^(jj+1)(2^jj+1):2^(jj+1)) = reshape(alpha(n1+1:n1+n2) 2^(jj+1) 2^(jj+1)-2^jj);
alphas_FDR(jj-j0+1) = {alpha};
% Solve the CS problem with FAR thresholding
a_0 = (0.4*Ndetail/Mdetail)/S;
tic
[alpha iters] = SolveStOMP(‘FastCSOperator‘ y Mdetail ‘FAR‘ a_0 S 1);
t_FAR = t_FAR + toc;
alpha_FAR((2^jj+1):2^(jj+1)1:2^jj) = reshape(alpha(1:n1) 2^(jj+1)-2^jj 2^jj);
alpha_FAR(1:2^(jj+1)(2^jj+1):2^(jj+1)) = reshape(alpha(n1+1:n1+n2) 2^(jj+1) 2^(jj+1)-2^jj);
alphas_FAR(jj-j0+1) = {alpha};
end
% Reconstruct and compute error
I_BP = ITWT2_PO(alpha_BP j0 qmf);
I_FDR = ITWT2_PO(alpha_FDR j0 qmf);
I_FAR = ITWT2_PO(alpha_FAR j0 qmf);
E_BP = norm(I0 - I_BP‘fro‘) / norm(I0‘fro‘);
E_FDR = norm(I0 - I_FDR‘fro‘) / norm(I0‘fro‘);
E_FAR = norm(I0 - I_FAR‘fro‘) / norm(I0‘fro‘);
save DataFig14.mat Narr N_CS j0 j1 I0 I_BP I_FDR I_FAR ...
t_BP t_FAR t_FDR E_BP E_FAR E_FDR thetas alphas_BP alphas_FDR alphas_FAR
%
% Copyright (c) 2006. D
属性 大小 日期 时间 名称
----------- --------- ---------- ----- ----
文件 860 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\aconv.m
文件 185727 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\barbara.png
文件 5531 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\BCS_fast_rvm.m
文件 177762 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\boat.png
文件 262750 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\cameraman.tif
文件 226 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\DataFig08b_random.mat
文件 606 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\DownDyadHi.m
文件 585 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\DownDyadLo.m
文件 4795 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\example.asv
文件 4793 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\example.m
文件 7246 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\FastLaplace.m
文件 1106 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\FWT2_PO.m
文件 3263 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\GenDataFig08b_random.m
文件 777 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\iconv.m
文件 1212 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\IWT2_PO.m
文件 466 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\lshift.m
文件 10081 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\MakeONFilter.m
文件 262750 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mandril_gray.tif
文件 591 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\MirrorFilt.m
文件 264168 2010-05-27 22:30 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mondrian(2)\Mondrian.tif
..A..H. 25088 2010-05-27 22:30 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mondrian(2)\Thumbs.db
文件 29209 2010-05-27 22:30 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mondrian(2)\unti
文件 42458 2010-05-27 22:30 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mondrian(2)\unti
文件 30465 2010-05-27 22:30 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mondrian(2)\unti
文件 34847 2010-05-27 22:30 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mondrian(2)\unti
文件 42806 2010-05-27 22:30 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mondrian(2)\unti
文件 38827 2010-05-27 22:30 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\mondrian(2)\unti
文件 54596 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\pdco.m
文件 10186 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\pdcoSet.m
文件 3824 2010-05-27 22:29 fusion_for CS\fastlaplacemorian\FastLaplace mondrian\phantom_256.tif
............此处省略71个文件信息
- 上一篇:基于改进的人工势场法的机器人路径规划
- 下一篇:cacfar单元平均恒虚警检测方法
评论
共有 条评论