How to use the find function discover the row and column [pixel coordinates] of all the blue pixels per column?

I’m having problems using the find function to discover the row and column [pixel coordinates] of the blue pixels.

What I want to do is find the first and last blue pixels these will correspond to the boundaries of the lumen where I want to paint the range between these first and last boundaries in blue, per column. But in some column values I have the bifurcation of the vessel, and here again I want to find the first and last blue pixel of the bifurcation (between the first and last blue pixels found for lumen) and paint the range between these values in black.

I was able to find the first and last blue pixels using the code below (but without using the row value jj), only with movingRegistered(:,zz,3)==).

1. First I don’t understand why it only works with (: )
2. But using the same logic to find the first and last bifurcation values between the values found, appears the error:Subscripted assignment dimension mismatch

I tried to make :

(movingRegistered- 499x631x3 uint8
Bif_firt(1x631)
Bif_last(1x631)

bif_first(1,zz)=find((movingRegistered(jj,zz,1)==0 & ...
                      movingRegistered(jj,zz,2)==0 & ...
                      movingRegistered(jj,zz,3)==255),1,'first'); % the first blue row for that column
bif_last(1,zz)=find((movingRegistered(jj,zz,1)==0 & ...
                     movingRegistered(jj,zz,2)==0 & ...
                     movingRegistered(jj,zz,3)==255),1,'last'); %the last blue row for that column

But it only appears 1 in every column.

        3-Does anybody have any suggestion?
      4- I tried to paint the image with blue and the rest with black but my figure (13) appears the same as before, what I’m doing wrong? Please see my code below:

%%Find the countour indexes and paint the lumen in blue  %%%%%%
lumen_first=zeros(1,size(movingRegistered,2));
lumen_last=zeros(1,size(movingRegistered,2));
bif_first=zeros(1,size(movingRegistered,2));
bif_last=zeros(1,size(movingRegistered,2));
lumen=false;
for zz=1:1:size(movingRegistered,2);%columns
  for jj=1:size(movingRegistered,1);%rows
    if(movingRegistered(jj,zz,1)==0 &&  ...
       movingRegistered(jj,zz,2)==0 &&  ...
       movingRegistered(jj,zz,3)==255)
      lumen=true;
      lumen_first(1,zz)=find((movingRegistered(:,zz,1)==0 & ...
                              movingRegistered(:,zz,2)==0 & ...
                              movingRegistered(:,zz,3)==255),1,'first'); % the first blue row for that column
      lumen_last(1,zz)=find((movingRegistered(:,zz,1)==0 & ...
                             movingRegistered(:,zz,2)==0  & ...
                             movingRegistered(:,zz,3)==255),1,'last'); %the last blue row for that column
    end
  end
end
for zz=1:1:size(movingRegistered,2);%columns
  for jj=lumen_first(1,zz)+4:lumen_last(1,zz)-1;%rows
    bifurcacao=false;
    if (movingRegistered(jj,zz,1)==0 && ... 
        movingRegistered(jj,zz,2)==0 && ...
        movingRegistered(jj,zz,3)==255)
      bifurcacao=true;    %existis bifurcation
      bif_first(1,zz)=find((movingRegistered(jj,zz,1)==0 & ...
                            movingRegistered(jj,zz,2)==0 & ...
                            movingRegistered(jj,zz,3)==255),1,'first'); % the first blue row for that column
      bif_last(1,zz)=find((movingRegistered(jj,zz,1)==0 & ...
                           movingRegistered(jj,zz,2)==0 & ...
                           movingRegistered(jj,zz,3)==255),1,'last');  %the last blue row for that column
    end
  end
end
rgbImage_blue=movingRegistered;
%%Paint in blue the lumen region
%Verify if where vector lumen_first has 0 values  
zmin_lumen=find(lumen_first(:),1,'first');%column value
zmax_lumen=find(lumen_last(:),1,'last');%column value
lumen_green=zeros(size(movingRegistered));
for z=zmin_lumen:zmax_lumen
  for j1=lumen_first(1,z):lumen_last(1,z);
    lumen_green(j1,z)=1;
  end
end
for zz=1:size(rgbImage_blue,2);
  for jj=size(rgbImage_blue,1);
    if lumen_green(jj,zz) ==1;
      rgbImage_blue(jj,zz,1)=0;
      rgbImage_blue(jj,zz,2)=0;
      rgbImage_blue(jj,zz,3)=255;
    else
      rgbImage_blue(jj,zz,1)=0;
      rgbImage_blue(jj,zz,2)=0;
      rgbImage_blue(jj,zz,3)=0;
    end
  end
  %     if bifurcacao,
  %         for j2=bif_first(1,zz):bif_last(1,zz)
  %             rgbImage_blue(j2,zz,1)=0;
  %             rgbImage_blue(j2,zz,2)=0;
  %             rgbImage_blue(j2,zz,3)=0;
  %         end

end
figure(13),imshow(rgbImage_blue);
<</matlabcentral/answers/uploaded_files/24482/movingRegistered.png>>

[dpb]

...I tried to make...

bif_first(1,zz)=find((movingRegistered(jj,zz,1)==0 & ...
                      movingRegistered(jj,zz,2)==0 & ...
                      movingRegistered(jj,zz,3)==255)

...But it only appears 1 in every column.

Because the there's only a single pixel at point jj,zz for each plane so if it matches the only location it can return is '1'.

find returns the location in array based on the size of the argument list; in your case actually on a pixel-by-pixel loop basis it would be the value jj itself that you're interested in when the match is made the first/last time.

One could code this via a loop and an if that you terminate when find the first and then a second loop from the end backwards for the last, but there's probably a much simpler way, depending on the image.

So, a question, first--is this a solid color so that the patch is of one color between two boundaries (of which there may be two pair) or is it varied so there are possibly many locations?

Alternatively, while I don't do image processing and don't have the toolbox, Image Analyst or other who do know the field and the Toolbox may have far better solutions...

[Image Analyst]

I'm having trouble visualizing this without an image to look at. And it looks like you're looking for pure blue, like computer graphics. I think registration works best with grayscale images, but it might work with line graphics too. I haven't used image registration much. Alex Taylor is the real expert (I think he wrote the MATLAB functions).

[dpb]

IA, I could envision a blob of pseudo-color and with the use of the words vessel and bifurcation I could think of something like a blood vessel so one had (say) the main branch and two (or more, possibly) smaller as one worked across an image. If that were the case and the false color was indeed pure blue, then I can see the logic in what Helena is trying to do. It's probably worth getting the clarification before spending time writing code, though.

Altho I will note one very big complexity reduction in the code that could make reading/writing it much simpler...if are going to write an if for comparison, then

all(x(i,j,:)==[0 0 255].');

is much simpler than the test for all three components individually. It would be better documented to write a vector for blue as well so could write

    all(x(i,j,:)==blue);

where

blue=[0 0 255].';

(Here's a place where it would be nice if Matlab had the equivalent of Fortran PARAMETER or similar for constants).

I've no clue as to whether this is even close or not; as you say, attaching an image would be of great benefit.

[Helena Henrques]

I see, we need I list/ array of values, with your answeer I tried this, but now occurs the problem of assignment dimensions mismatch any suggestion???

%Find the first and last rows per column to define the range of biffurcation to paint in black 
 col_bif=zeros(size(movingRegistered,1)-50,1);
 for zz=1:1:size(movingRegistered,2);%columns
     col_bif((lumen_last(1,zz)-1)-(lumen_first(1,zz)+4),zz)=movingRegistered((lumen_first(1,zz)+4:lumen_last(1,zz)-1),zz);%rows
     if (col_bif(:,zz,1)==0 && col_bif(:,zz,2)==0 && col_bif(:,zz,3)==255);
         bifurcacao=true;%existis bifurcation
         bif_first(1,zz)=find((col_bif(:,zz,1)==0 & col_bif(:,zz,2)==0 & col_bif(:,zz,3)==255),1,'first'); % the first blue row for that column
         bif_last(1,zz)=find((col_bif(:,zz,1)==0 & col_bif(:,zz,2)==0 & col_bif(:,zz,3)==255),1,'last'); %the last blue row for that column
     end
 end

[dpb]

OK, my guess wasn't too far off... :)

As said earlier, I'm not an image processing person by any stretch, but the way I'd tackle this from a Matlab standpoint would be by use of logicals and differencing to find the edges and then count those changes in sign to determine the number of edges and their locations.

Consider what happens as you traverse a column -- it's not then is for a few pixels, then isn't the rest of the way until the other side of the vessel at which time the same thing occurs. That's what the first and last do with find automagically for only one but the problem is you don't know whether that is/isn't the right answer.

OTOH, (I'll use x simply for brevity of the image), consider

ix=(x==blue);      % logical array of where blue pixels are
dx=[0;diff(ix)];   % the difference by row for all columns
nTops=sum(dx==1):  % the transitions _to_ blue
nBots=sum(dx==-1); % and _from_ blue...
% see if are consistent; if not, something's not right in image
if(nTops~=nBots), error('Mismatch in Top/Bottom'),end

Now, if Ntops==2, you know there's only one body with two boundaries and the extent is either the aforementioned find result for those columns or, since already have the info it's the first/last nonzero locations in ix.

If nTops>2 (and even is another sanity check) then for each pair it's the same locations -- the top and bottom are each successive alternate +1/-1 pair.

Attach the image file if you can and I'll bet you'll get a solution in...name that tune minutes... :)

I've gotta' run right now or I'd make up some examples.

[Helena Henrques]

Thanks dpb for your answeer,I try your advice to optimize the code but It doesn't seems to work: Error using == matrixdimensions must agree! Can you see my previous answeer, related to the find. I creacte a column per each z value, but the dimensions are different in every loop and appears that error.

lumen=false;
blue=[0 0 255].';

for zz=1:1:size(movingRegistered,2);%columns
    for jj=1:size(movingRegistered,1);%rows
        if (all(movingRegistered(jj,zz,:)==blue));
            lumen_first(1,zz)=find(all(movingRegistered(jj,zz,:)==blue),1,'first'); % the first blue row for that column

              lumen_last(1,zz)=find(all(movingRegistered(jj,zz,:)==blue),1,'last');%the last blue row for that column
          end
      end
  end

[Helena Henrques]

I want to find per column the first and last blue pixels and them find the other boundary ( first and last blue pixel rows) inside this rows achieved before.

I made the alterations based on the previous answeers, but now I have this error when it tries to find the bif_first and last: Subscripted assignment dimension mismatch. bif_first2(2,z)=find(dx(:,z)==1,2,'first');

What is wrong? Please see my code below:

rgbImage_blue=zeros(size(movingRegistered));

lumen_first=zeros(1,size(movingRegistered,2));
lumen_last=zeros(1,size(movingRegistered,2));
bif_first=zeros(1,size(movingRegistered,2));
bif_last=zeros(1,size(movingRegistered,2));
bif_last2=zeros(2,size(movingRegistered,2));
bif_first2=zeros(2,size(movingRegistered,2));

blue=cat(3,0,0,255);
ix=all(bsxfun(@eq,movingRegistered,blue),3);% find all the blue pixels, and put them at 1
% logical array of where blue pixels are
dx=[zeros(1,size(movingRegistered,2));diff(ix)]; % zeros to make the same number of columns and rows
% the difference by row for all columns
nTops=sum(dx==1);  % the transitions _to_ blue
nBots=sum(dx==-1); % and _from_ blue...
% see if are consistent; if not, something's not right in image
if(nTops~=nBots), error('Mismatch in Top/Bottom'),end

   for z=1:1:size(movingRegistered,2);
      if  nTops(1,z)==2; bifurcation=false;lumen=true; %only existis two boundaries no bifurcation
      lumen_first(1,z)=find(ix(:,z)==1,1,'first');
      lumen_last(1,z)=find(ix(:,z)==1,1,'last');
      end
      if nTops(1,z)>2;
      bifurcation=true;
      lumen_first(1,z)=find(ix(:,z)==1,1,'first');
      lumen_last(1,z)=find(ix(:,z)==1,1,'last');
      bif_first2(2,z)=find(dx(:,z)==1,2,'first');
      bif_first(1,z)=bif_first2(2,z);
      bif_last2(2,z)=find(dx(:,z)==1,2,'last');
      bif_last(1,z)=bif_last2(2,z);
      end
   end

[dpb]

Oh, yeah, I don't do 3D that much, so the 2D indexing doesn't work. I'll try to find the time later on to look at this a little more and work out the syntax problems in my approach. I know it'll work; I just have to work thru the syntax.

Is the outline really "true blue" of [0 0 255]? I pasted the image from the screen shot bitmap and saved it to a jpg for 'spearminting but there the colors aren't pure.

I'm wondering that if they really are, maybe you get the same result if simply look at the blue plane alone? Would there be any pixels of full 255 blue saturation also of anything but blue? If that were true would reduce the problem size by 2/3rds...plus only be a 2D instead of 3D problem.