How to apply a function to every element of a matrix

Hi, I have a RGB image which in MATLAB is in the form of a 3D matrix. The code:

    angle_dimension = linspace(0,180,3);
    cashew_RGB = imread('White Wholes 180.jpg');
    cashew_BW = img_BW(cashew_RGB);
    cashew_stats = regionprops(cashew_BW,'BoundingBox');
    for i = 1:numel(cashew_stats)
      cashew_single = imcrop(cashew_RGB,cashew_stats(i).BoundingBox);
        cashew_single_BW = img_BW(cashew_single);
      dim = imFeretDiameter(cashew_single_BW,angle_dimension);
    end

The functions used are not really important. I want to know how I can eliminate the for loop. It is possible in my project that

numel(cashew_stats)

may return a large number and the for loop then would take a lot of time to execute. It would be immensely helpful if you could give me a technique to eliminate the for loop used. I have attached the image that I have used in the code. Thank You!

Answer 1

[Image Analyst]

You don't have a loop over the number of elements (which is 3 times the number of pixels) in the image! You have a loop over the regions you found. And you can't use Bounding Box because other cashews could poke into the bounding box. Actually you don't even use regionprops() at all! You need to use ismember() and then bwboundaries(). Then there is no feret diameter function in MATLAB so you have to use my attached program to get the two farthest points.

[labeledImage, numRegions] = bwlabel(cashew_BW);
cashew_stats = regionprops(labeledImage,'BoundingBox');
for k = 1 : numRegions
    % Get binary image of just this one cashew.
    thisCashew = ismember(labeledImage, k) > 0;
    % Get its boundaries
    boundaries = bwboundaries(thisCashew);
    x = boundaries(:, 2);
    y = boundaries(:, 1);
    feretDiameters(k) = GetFeretDiameters(x, y); % See my attached demo for this.
end

[Nikhil Hegde]

I'm in awe of your knowledge Image Analyst. I'm still a novice at MATLAB. And I know the for loop was not going over the pixels of the image. I just thought maybe it was possible to eliminate it. As number of cashews per sample increase, the loop increases too. But thanks a lot! Appreciate it!

[Image Analyst]

Yes, but for loops are no problem when the number of iterations is less than a few million, and I doubt you'll have that many cashews in your image. Even when the number of iterations is hundreds of millions, it's usually the memory access on a huge array that is causing the slow down since I can have a hundred million iterations where I don't access any other variables and the time to complete the loop is only 0.3 seconds (I actually just tried it now). Now, if inside that loop I was accessing the (i,j) element of an array with hundreds of millions of elements, then that would be slower, but the for loop itself is not the problem.

[Nikhil Hegde]

That's true. I checked the execution time as well and is around 0.3 seconds. I'm implementing this as a real time system so it's my utmost priority to keep the execution time as low as possible. I read somewhere that for loops slow down execution and hence this question. Thanks a lot for your response!