richardson-luck 迭代盲复原算法-MATLAB

clc clear all; %PSF-点扩散函数的尺寸与图像大小一致，不用考虑点扩散函数的尺寸。这是RL算法的一个优点。2015年12月12日。 f = double(imread('sat.tif')); h = gauskenerl(f); g = dftconv(f,h); colormap Jet; subplot(221), imagesc(g),title(sprintf('blurred image'));axis equal, axis tight,axis off; [L hh] = hsize(h); subplot(222), surf(1:L,1:L,hh);title(sprintf('origin PSF ')); %RL-IBD alogrithm h = h0_est(g); ImgLatent = g; tic for k = 1:30 for i = 1:7 %%%%%%%%%%%%%%%%%%% 估计PSF %%%%%%%%%%%%%%%%%%%%%%%%%%%% %第1步 ImgReblur0 = dftconv(ImgLatent,h); %第2步 relative_blur0 = g./ImgReblur0; %第3步 error_est0 = dftconv(relative_blur0,rot90(ImgLatent,2)); %第4步 h = 1/sum(ImgLatent(:)) * (h .* error_est0) ; end for j = 1:1 %%%%%%%%%%%%%%%%%%% 估计图像f %%%%%%%%%%%%%%%%%%%%%%%%%%%% %第1步 ImgReblur = dftconv(ImgLatent,h); %第2步 relative_blur = g./ImgReblur; %第3步 error_est = dftconv(relative_blur,rot90(h,2)); %第4步 ImgLatent = ImgLatent .* error_est; end subplot(223), imagesc(ImgLatent),title(sprintf('estimated image x %d', k));axis equal, axis tight,axis off; [L hh] = hsize(h); subplot(224), surf(1:L,1:L,hh);title(sprintf('estimated PSF x %d', k)); drawnow end toc %%尺寸：189*189，迭代次数（30,7,1），共使用42秒。∴ 迭代0.175秒/次。