How do I select a window of pixels around a pixel?

Hi

I am navigating through each pixel in an image(grayscale binary image) using a for loop. For each pixel, I need to select a window of 5*5 pixels around it and perform certain operations on it. If the operation is successful, i need to move to the next pixel, if not, I need to increase the size of the window till it is successful!

Any help will be appreciated. Thanks in advance.

Answer 1

[AJ von Alt]

You can access multiple elements of a matrix using a vector of indices. E.g.:

    B = A( (i - 1 : i + 1) , (i-1 : i + 1) )

Consider a 3 x 3 matrix A:

    A = [11 22 33 ; 
         44 55 66 ; 
         77 88 99 ];

The following command will get the sub-matrix made up of the first two rows and columns of A and store that sub-matrix in B.

    B = A( 1:2 , 1:2 )

B =

      11    22
      44    55

The following code is an example implementation tracking the number of times the number of windows evaluated before the success condition is met.

img               = double( imread( 'trees.tif') );
[ nRows , nCols ] = size( img );
result            = zeros( nRows , nCols );
halfWinSize       = 2; 
threshold         = 50;

for rI = 1:nRows
    for cI = 1:nCols
        % These indices specify the start and stop indices of the
        % window. The window will go from row rILo to rIHi amd column
        % cILo to cIHi
        rILo               = rI - halfWinSize; % lower row index of window
        rIHi               = rI + halfWinSize; % upper row index of window
        cILo               = cI - halfWinSize; % lower column index of window
        cIHi               = cI + halfWinSize; % upper column index of window
        flagConditionMet = false ;
        count = 1;

          % while success criteria unmet
          while( ~ flagConditionMet )
          % It is important to make sure that your window indices are valid
          % indices before pulling out the window.            
              if( rILo >= 1 && rIHi <= nRows && cILo >= 1 && cIHi <= nCols )

                  % Pull out the window around (rI , cI)
                  imgWindow = img( (rILo : rIHi) , (cILo : cIHi) ); 

                  %perform an operation
                  minVal = min( imgWindow(:) );

                  %check if successful
                  flagConditionMet = minVal < threshold;

                  %increase window bounds if unsuccessful   
                  if( ~ flagConditionMet )                  
                      rILo = rILo - 1;
                      rIHi = rIHi + 1;
                      cILo = cILo - 1;
                      cIHi = cIHi + 1;
                      count = count + 1;
                  else
                      %store some result if successful
                      result( rI , cI ) = count;                    
                  end

              else 
                  % out of bounds break out of while loop
                  break; 
              end
          end
      end
  end

figure;
imshow(result./max( result(:) ))

[Rachel]

Thanks a lot!

I actually wanted to implement a rectangular sliding window around a pixel. Will this work for that as well?

Answer 2

[deep tank]

Hello AJ von Alt and Rachel

Can you help me to develop matlab code for equation highlighted as RED (page 3 of 5) in attached file.....I feel it is similar to above query.

Thanks in advanced.

Regards, Dipesh