How to scan an image from a specified Row?

So below is a function Im using to locate the beginning and end of dark regions on a greyscale image:

function [Ssb Seb]=get_Black_edges(A)
[r c]=size(A);
Ssb=[];  %start Black 
Seb=[];  %end Black
for k=1:1:c
   x=A(:,k);
   a=find(x<26);
mid=floor(mean(a));
i=mid;
while(x(i-1) <26 | x(i-2) <26)
    i=i-1;
end
sb=i;
Ssb=[Ssb ; sb];
i=mid;
while(x(i+1) <26 | x(i+2) <26)
    i=i+1;
end
eb=i;
Seb=[Seb ; eb];
end
return

Say my image is 886x1024, how to I adjust the code so that the scanning ignores the first 40 rows and only scans the remaining 846?

[Aatheeswaran M]

whats A in the function?

[Stelios Fanourakis]

I get the error

Subscript indices must either be real positive integers or logicals.
Error in efr>get_Black_edges (line 16)
while (x(i-1) <26 | x(i-2) <26)
Error in efr (line 3)
get_Black_edges(A); 
>> 

Answer 1

[Guillaume]

Your code would be so much easier to understand if you'd use meaningful variable names. It's particularly confusing that you're using x to represent pixel intensities when it's usually used to represent pixel coordinates.

In any case, why can't you just pass the portion of the image of interest to your function? That is instead of:

[blackstartingrows, blackendingrows] = get_Black_edges(someimage);

do

[blackstartingrows, blackendingrows] = get_Black_edges(someimage(41:end, :));
blackstartingrows = blackstartingrows + 40; %readjust to original image coordinates
blackendingrows = blackendingrows + 40;

[Guillaume]

Note that your code will error if all the pixels at the top or the bottom of a row are below 26, since your first while loop can try to access pixels on row 0, -1, and the second on row height+1, height+2.

Secondly, are you sure that the condition you want for scanning up or down is | (or) and not & (and) ?

[Guillaume]

Going back to my comment about using meaningful variable names, don't you find the following a lot more self documenting and thus easier to understand:

function [blackstartingrows, blackendingrows] = get_Black_edges(img)
   [height, width] = size(img);
   blackstartingrows = zeros(1, width); %better preallocate as well, rather than grow
   blackendingrows = zeros(1, width);
   for column = 1:width
      columnpixels = img(:, column);
      rowsbelowthresh = find(columnpixels < 26);
      meanrow = floor(mean(rowsbelowtresh));
      %the whole while loop can be replaced with just one line:
      blackstartrow = find([Inf columnspixel(1:meanrow)] > 26 & [Inf Inf columnspixel(1:meanrow-1) > 26], 1, 'last') + 1; 
      blackstartingrows(column) = blackstartrow;
      %same with other loop, it's just:
      blackendrow = find([columnspixel(meanrow+1:end) Inf] > 26 & [columnspixel(meanrow+2:end) Inf Inf], 1, 'first') - 1;
      blackendingrows(column) = blackendrow;
   end
end

[Sean]

Hi thanks I really appreciate the feedback, I slightly adjusted the code above due to some typos:

function [blackstartingrows, blackendingrows] = get_Black_edges3(A)
   [height, width] = size(A);
   blackstartingrows = zeros(1, width); %better preallocate as well, rather than grow
   blackendingrows = zeros(1, width);
   for column = 1:width
      columnpixels = A(:, column);
      rowsbelowthresh = find(columnpixels < 26);
      meanrow = floor(mean(rowsbelowthresh));
      %the whole while loop can be replaced with just one line:
      blackstartrow = find([Inf columnpixels(1:meanrow)] > 26 & [Inf Inf columnpixels(1:meanrow-1) > 26], 1, 'last') + 1; 
      blackstartingrows(column) = blackstartrow;
      %same with other loop, it's just:
      blackendrow = find([columnpixels(meanrow+1:end) Inf] > 26 & [columnpixels(meanrow+2:end) Inf Inf], 1, 'first') - 1;
      blackendingrows(column) = blackendrow;
   end
end

columnspixel and columnpixels as an example, I ran the code however and the following error came up:

Error using horzcat
Dimensions of matrices being concatenated are not consistent.

Error in get_Black_edges3 (line 12)
      blackstartrow = find([Inf columnpixels(1:meanrow)] > 26 & [Inf Inf columnpixels(1:meanrow-1) > 26], 1, 'last') + 1;

Is there a problem with the adjustments I made?

Again thanks for the help! and yes I can see through the methodology of better use of variable names! much easier to follow!

[Image Analyst]

Sean, please read this and fix your code.

[Guillaume]

Sean, yes, I didn't test the code, just typed it in my answer, so expect some typos and minor bugs.

I forgot that columnpixels was a column (!). Therefore you need a vertical concatenation (a.k.a semicolons), that is:

blackstartrow = find([Inf; columnpixels(1:meanrow)] > 26 & [Inf; Inf; columnpixels(1:meanrow-1) > 26], 1, 'last') + 1;

and

blackendrow = find([columnpixels(meanrow+1:end); Inf] > 26 & [columnpixels(meanrow+2:end); Inf; Inf], 1, 'first') - 1;

[Sean]

Hi, I got the code working! Thanks a lot again for the help. Much appreciated!