How to vectorize for..loop with nested "if" and "break" statements

Dear colleagues, I am trying to vectorize the following for..loop in my matlab code:

for c=Cmin:Cmax % Cmin, Cmax are columns indices
    for r=Rmin:Rmax % Rmin, Rmax are rows indices
        if(img1(r, c)==1) % img1 is a binary image
            x1 = r;
            y1 = c;
            break;
        end
    end
end

The problem I am facing is the inner "if" and "statement" to be included in the vectorized code. I have followed several vectorization techniques, but I haven't happened to see one that include nested conditions. Any idea please. Thank you!

[dpb]

'Pends on what needs to be done inside the loop besides what is (presumably) demo code rather than real as the end result of the above snippet will simply be the last set that meet the criterion; all the earlier are lost/overwritten. Is saving an array of those locations sufficient for use later or is something else needing done at the time of the discovery???

[Jos (10584)]

it is the first occurrence , due to the break statement

[Kevin Claytor]

Are you actually trying to determine if this is a binary image? Because there are better ways of doing that, and your code snippet above would risk misclassifying any integer-valued image as a binary image.

[dpb]

"_it is the first occurrence , due to the break statement"_

The break only terminates the inner loop; it'll then go on and start over the outer loop. Since x1, y1 are each a single variable, you'll overwrite the first column location with the second, then the third, leaving at the end only the last column first row as the one and only pair of values retained.

Answer 1

[Orion]

Hi, it seems you're trying to do something like

% initialize indices
Rmin = 3;
Rmax = 8;
Cmin = 2;
Cmax = 7;

% initialize a zero matrix with some values to one.
img = zeros(10);
img(4,5) = 1; % inside Rmin/Rmax, Cmin/Cmax
img(3,7) = 1; % inside Rmin/Rmax, Cmin/Cmax
img(3,8) = 1; % OUTSIDE Rmin/Rmax, Cmin/Cmax => won't be in the result

% get all the points equal to 1 in the Rmin/Rmax, Cmin/Cmax submatrix
[allx,ally] = find(img(Rmin:Rmax,Cmin:Cmax)==1);

% correct the indices to fit those in the whole matrix.
allx = allx + Rmin - 1;
ally = ally + Cmin - 1;

% get the first component :
x1 = allx(1);
y1 = ally(1);

[Baraka Maiseli]

Thank you Orion for a nice comment. Your code works well; I also tried a short form by Jos (next comment), which also produce a correct result.

Answer 2

[Jos (10584)]

[x1,y1] = find(img1(Rmin:Rmax,Cmin:Cmax)==1,1,'first')

and perhaps correct for the offset

x1 = x1 + Rmin - 1 ...

[Baraka Maiseli]

Jos, I never knew such a lengthy for..loop could be put in such a simple two-line code (I remember one of my lecturer told us in a class, "think more, write less"). Your code works very well. Now, I am trying to get another similar pair (x2,y2), which is somewhere from the end of image, img1, as shown in the illustration. Let me try to apply your idea to get the point.

[Baraka Maiseli]

Got it! As the positions of 1's are on the borders of the bounding box, with a little offset (observed from experiments), I fixed Rmin and Cmin and computed appropriate indices:

[x1,y1] = find(img1(Rmin:Rmax,Cmin+1)==1,1,'first'); 
          % Also, we may use [x1,y1] = find(img1(Rmin:Rmax,Cmin:Cmax)==1,1,'first');
          % for searching (x1, y1) within the bounding box
x1 = x1 + Rmin; % Correction for offset
y1 = y1 + Cmin; % Correction for offset

[x2,y2] = find((img1(Rmin+1,Cmin:Cmax))==1,1,'first');
          % Also, we may use [x2,y2] = find(img1(Rmin:Rmax,Cmin:Cmax)==1,1,'first');
          % for searching (x2, y2) within the bounding box
x2 = x2 + Rmin; % Correction for offset
y2 = y2 + Cmin; % Correction for offset