"Gradient Domain Guided Image Filtering"论文中边缘权重函数matlab实现代码

本文为转载，原博客地址为：http://blog.csdn.net/majinlei121/article/details/50458218

"Kou F, Chen W, Wen C, et al. Gradient Domain Guided Image Filtering[J]. Image Processing, IEEE Transactions on, 2015, 24(11): 4528-4539."论文中权重函数matlab实现的代码。

 clear all;

 g = double(imread('tulips.png'))/255.0;
 [m, n, d]=size(g);
 if(size(g,3)==3)
     g=rgb2gray(g);
 end

 %% weight 1
 % epsilon=(0.001*(max(g(:))-min(g(:))))^2;
 % r=1;
 %
 % N = boxfilter(ones(m, n), r); % the size of each local patch; N=(2r+1)^2 except for boundary pixels.
 % mean_I = boxfilter(g, r) ./ N;
 % mean_II = boxfilter(g.*g, r) ./ N;
 % var_I = mean_II - mean_I .* mean_I;
 %
 % weight1=(var_I+epsilon)/(mean(var_I(:))+epsilon);

 %% weight 2
 epsilon=(0.001*(max(g(:))-min(g(:))))^2;
 r1=1;
 r2=16;

 N1 = boxfilter(ones(m, n), r1); % the size of each local patch; N=(2r+1)^2 except for boundary pixels.
 mean_I1 = boxfilter(g, r1) ./ N1;
 mean_II1 = boxfilter(g.*g, r1) ./ N1;
 var_I1 = mean_II1 - mean_I1 .* mean_I1;
 N2 = boxfilter(ones(m, n), r2); % the size of each local patch; N=(2r+1)^2 except for boundary pixels.
 mean_I2 = boxfilter(g, r2) ./ N2;
 mean_II2 = boxfilter(g.*g, r2) ./ N2;
 var_I2 = mean_II2 - mean_I2 .*mean_I2;

 chi_I=sqrt(abs(var_I1.*var_I2));
 weight1=(var_I1+epsilon)/(mean(var_I1(:))+epsilon);
 weight2=(chi_I+epsilon)/(mean(chi_I(:))+epsilon);
 weight1=mat2gray(weight1);
 weight2=mat2gray(weight2);

 figure;imshow(weight1);
 colormap('pink');
 figure;imshow(weight2);
 colormap('pink');

下面是效果：