图像去雾--包含soft matting的matlab代码

img_name='333.jpg';
% 原始图像
I=double(imread(img_name))/255;

% 获取图像大小

[h,w,c]=size(I);
win_size = 7;
img_size=w*h;
dehaze=zeros(img_size*c,1);
dehaze=reshape(dehaze,h,w,c);

figure, imshow(I);
win_dark=zeros(img_size ,1);

for cc=1:img_size
   win_dark(cc)=1;
end

win_dark=reshape(win_dark,h,w);
%计算分块darkchannel
 for j=1+win_size:w-win_size
    for i=win_size+1:h-win_size
        m_pos_min = min(I(i,j,:));
        for n=j-win_size:j+win_size
            for m=i-win_size:i+win_size
                if(win_dark(m,n)>m_pos_min)
                    win_dark(m,n)=m_pos_min;
                end
            end
        end
      
    end
 end
 
 figure, imshow(win_dark);
 
 for cc=1:img_size
   win_dark(cc)=1-win_dark(cc);
 end
 %选定精确dark value坐标
win_b = zeros(img_size,1);
 
for ci=1:h
    for cj=1:w
        if(rem(ci-8,15)<1)
            if(rem(cj-8,15)<1)           %这么做的目的是-在全矩阵中得到稀疏点--均匀分布在矩阵中
                win_b(ci*w+cj)=win_dark(ci*w+cj);
            end
        end
      
    end
end
 
%显示分块darkchannel
%figure, imshow(win_dark);
neb_size = 9;
win_size = 1;
epsilon = 0.0000001;
%指定矩阵形状
indsM=reshape([1:img_size],h,w);
%计算矩阵L
tlen = img_size*neb_size^2;
row_inds=zeros(tlen ,1);
col_inds=zeros(tlen,1);
vals=zeros(tlen,1);
len=0;
for j=1+win_size:w-win_size
    for i=win_size+1:h-win_size
        if(rem(ci-8,15)<1)
            if(rem(cj-8,15)<1)
                continue;
            end
        end
      win_inds=indsM(i-win_size:i+win_size,j-win_size:j+win_size);
      win_inds=win_inds(:);%列显示
      winI=I(i-win_size:i+win_size,j-win_size:j+win_size,:);
      winI=reshape(winI,neb_size,c); %三个通道被拉平成为一个二维矩阵 3*9
      win_mu=mean(winI,1)';  %求每一列的均值 如果第二个参数为2 则为求每一行的均值  //矩阵变向量
      win_var=inv(winI'*winI/neb_size-win_mu*win_mu' +epsilon/neb_size*eye(c)); %求方差

      winI=winI-repmat(win_mu',neb_size,1);%求离差
      tvals=(1+winI*win_var*winI')/neb_size;% 求论文所指的矩阵L

      row_inds(1+len:neb_size^2+len)=reshape(repmat(win_inds,1,neb_size),...
                                             neb_size^2,1);
      col_inds(1+len:neb_size^2+len)=reshape(repmat(win_inds',neb_size,1),...
                                             neb_size^2,1);
      vals(1+len:neb_size^2+len)=tvals(:);
      len=len+neb_size^2;
    end
end

 

 

vals=vals(1:len);
row_inds=row_inds(1:len);
col_inds=col_inds(1:len);
%创建稀疏矩阵
A=sparse(row_inds,col_inds,vals,img_size,img_size);
%求行的总和 sumA为列向量
sumA=sum(A,2);
%spdiags(sumA(:),0,img_size,img_size) 创建img_size大小的稀疏矩阵其元素是sumA中的列元素放在由0指定的对角线位置上。
A=spdiags(sumA(:),0,img_size,img_size)-A;

  %创建稀疏矩阵
  D=spdiags(win_b(:),0,img_size,img_size);
  lambda=1;
  x=(A+lambda*D)/(lambda*win_b(:).*win_b(:));
 
   %去掉0-1范围以外的数
   alpha=max(min(reshape(x,h,w),1),0);
 

figure, imshow(alpha);
A=220/255; %大气光没有去计算
%去雾
      
for i=1:c
    for j=1:h
        for l=1:w
            dehaze(j,l,i)=(I(j,l,i)-A)/alpha(j,l)+A;
        end
    end
end

figure, imshow(dehaze);

 

 

其中对于L矩阵的求解还不是很了解，那个win_b为什么要那样取？   看了原文A Closed Form Solution to Natural Image Matting后，还是不大了解，有谁如果很了解，并想帮忙的话，联系我啊，不慎感激。