可能去霧代碼

1、可能去霧代碼img_name='333.jpg'% 原始圖像I=double(imread(img_name)/255;% 獲取圖像大小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;endwin_dark=reshape(win_dark,h,w);%計(jì)算

2、分塊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

3、 m=i-win_size:i+win_size                if(win_dark(m,n)>m_pos_min)                    win_dark(m,n)=m_pos_min;  

4、0;             end            end        end           end end  figure, imshow(win_da

5、rk);  for cc=1:img_size   win_dark(cc)=1-win_dark(cc); end %選定精確dark value坐標(biāo)win_b = zeros(img_size,1); for ci=1:h    for cj=1:w        if(rem(ci-8,15)<1)          

6、;  if(rem(cj-8,15)<1)           %這么做的目的是-在全矩陣中得到稀疏點(diǎn)-均勻分布在矩陣中                win_b(ci*w+cj)=win_dark(ci*w+cj);         

7、0;  end        end           endend %顯示分塊darkchannel%figure, imshow(win_dark);neb_size = 9;win_size = 1;epsilon = 0.0000001;%指定矩陣形狀indsM=reshape(1:img_size,h,w);%計(jì)算矩陣Ltlen = img_size*neb_size2;row_inds=zer

8、os(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)  

9、60;             continue;            end        end      win_inds=indsM(i-win_size:i+win_size,j-win_size:j+win_size); &#

10、160;    win_inds=win_inds(:);%列顯示      winI=I(i-win_size:i+win_size,j-win_size:j+win_size,:);      winI=reshape(winI,neb_size,c); %三個(gè)通道被拉平成為一個(gè)二維矩陣 3*9      win_mu=mean(winI,1)'  %求每一列的均值 如果第二個(gè)參數(shù)為2 則為求每

11、一行的均值  /矩陣變向量      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);%求離差      tvals=(1+winI*win_var*winI')/neb_size;% 求論文所指的矩陣L 

12、     row_inds(1+len:neb_size2+len)=reshape(repmat(win_inds,1,neb_size),.                                  

13、0;          neb_size2,1);      col_inds(1+len:neb_size2+len)=reshape(repmat(win_inds',neb_size,1),.                     &

14、#160;                       neb_size2,1);      vals(1+len:neb_size2+len)=tvals(:);      len=len+neb_size2;    endend 

15、60;vals=vals(1:len);row_inds=row_inds(1:len);col_inds=col_inds(1:len);%創(chuàng)建稀疏矩陣A=sparse(row_inds,col_inds,vals,img_size,img_size);%求行的總和 sumA為列向量sumA=sum(A,2);%spdiags(sumA(:),0,img_size,img_size) 創(chuàng)建img_size大小的稀疏矩陣其元素是sumA中的列元素放在由0指定的對(duì)角線位置上。A=spdiags(sumA(:),0,img_size,img_size)-A;  %創(chuàng)建稀疏矩陣 

16、 D=spdiags(win_b(:),0,img_size,img_size);  lambda=1;  x=(A+lambda*D)/(lambda*win_b(:).*win_b(:);    %去掉0-1范圍以外的數(shù)   alpha=max(min(reshape(x,h,w),1),0); figure, imshow(alpha);A=220/255; %大氣光沒有去計(jì)算%去霧       for i=1:c  &

17、#160; for j=1:h        for l=1:w            dehaze(j,l,i)=(I(j,l,i)-A)/alpha(j,l)+A;        end    endendfigure, imshow(dehaze);暗通道圖片參考算法 for (Y = 0, DarkPt = D

18、arkChannel; Y < Height; Y+) ImgPt = Scan0 + Y * Stride; for (X = 0; X < Width; X+) Min = *ImgPt; if (Min > *(ImgPt + 1) Min = *(ImgPt + 1); if (Min > *(ImgPt + 2) Min = *(ImgPt + 2); *DarkPt = Min; ImgPt += 3; DarkPt+; MinFilter(DarkChannel, Width, Height, Radius);去霧代碼三close allclcrgb=imread('image processingfogfog.jpg');% 對(duì)飽和度與亮度進(jìn)行hist