Hi, I am applying an algorithm on an Image which is so big 7001x7851. I am using many loops according to the algorithm and it takes much time to get the result. I am trying to use block processing but do not know how to get the 'fun'. here is my code

Tb1= imread([datapath71,datafile71]);
Fb1= imread([datapath51,datafile51]);
[r,c] = size(Tb1);  
[m,n] = size(Fb1); 
if r ~= m || c ~= n  
Fb1 = imresize(Fb1, [r,c]); 
end
figure;imshow(Fb1);title('Actual');
figure(2);imshow(Tb1);title('Target');

Tb1=im2double(Tb1);
Fb1=im2double(Fb1);
 [r,c]=size(Tb1);
 no = nnz(~Tb1);
fprintf('Total no of zeros = %g.\n',no);

no1=0; % no of filled gaps

% win_s='Enter the window size';
W = 9;

   Mt= zeros(size(Tb1));
   Mf= zeros(size(Fb1));
   comv=zeros(size(Tb1));
  for i=W:r-(W-1)
      for j=W:c-(W-1)

        if Tb1(i,j)==0
           com=0;St=0;Sf=0;

              for k1=i-(W-1):i+(W-1)
                  for k2=j-(W-1):j+(W-1)

               if (Tb1(k1,k2)>0 && Fb1(k1,k2)>0)
                      St=St+Tb1(k1,k2);
                      Sf=Sf+Fb1(k1,k2);
                         com=com+1;
               end;
                 end;
             end;
          Mt(i,j)=St/com;
          Mf(i,j)=Sf/com;
          comv(i,j)=com;

       end;
       end;
  end;

   New_dn= zeros(size(Tb1));
  for i=W:r-(W-1)
      for j=W:c-(W-1)

           if Tb1(i,j)==0
               SDt=0;SDf=0;

                 for k1=i-(W-1):i+(W-1)
                     for k2=j-(W-1):j+(W-1)
                         if (Tb1(k1,k2)>0 && Fb1(k1,k2)>0)

                  SDt=SDt+(Tb1(k1,k2)-Mt(i,j))^2;
                  SDf=SDf+(Fb1(k1,k2)-Mf(i,j))^2;
                         end;
                    end;
                 end;
         SD_t=SDt/(comv(i,j)-1);
         SD_f=SDf/(comv(i,j)-1);
          G=SD_t/SD_f;

               if (G>3||G<0.3)
                  G=1;
                end;

            Bi=Mt(i,j)-Mf(i,j)*G;

            ND=Fb1(i,j)*G+Bi;

              New_dn(i,j)=ND;
              no1=no1+1;

      else
          New_dn(i,j)=Tb1(i,j);
          end;
       end;

   end;

figure(3);imshow(New_dn);
imwrite(New_dn,'LLHM_b1_4.tif');

   fprintf('Total no of filled gaps in Band1= %g.\n',no1);

[Walter Roberson]

What is the code intended to do ?

Answer 1

[Image Analyst]

I don't know what this does but it's not processing in blocks. It's scanning a window a pixel at a time. With only a short glance, it looks like the whole quadruple for loop could be replaced by 3 or 4 lines of code, one to make a binary image of Tb1 and Fb1 are > 0, then another to set Tb1 and Fb1 equal to zero where the binary image is false, then a call to conv2().