How to detect each color square of an image to crop as individual images

I need to detect each square region with some kind of rectangle detection, I looked around other questions and there are some solutions with gray images but I couldn't find anything to detect with colorful images.

Ideally I need to detect each square, crop that region as another image, then perhaps add that image to an array, apply this detection, cropping and adding to the array, for each color region (16 regions for my case) for further process with each one of them.

How can I do this? Any suggestions are greatly appreciated.

[KSSV]

Attach your original image....

[Emre Havan]

This is the original image, it was noised and I pre-processed it to remove noise.

[DGM]

See also:

https://www.mathworks.com/matlabcentral/answers/1664934-returning-an-array-of-colors-from-a-double-image#answer_912294

https://www.mathworks.com/matlabcentral/answers/517876-colour-pattern-reading-using-only-lab-or-hsv

Answer 1

[Image Analyst]

Really easy. Here are the steps (untested):

% Convert RGB to HSV color space. 
hsvImage = rgb2hsv(rgbImage);
% Extract the Saturation channel and threshold it at, say, 0.25 
mask = hsvImage(:,:,2) > 0.25;
% extract the 16 largest blobs
mask = bwareafilt(mask, 16);
% Get bounding boxes
props = regionprops(mask, 'BoundingBox') 
% Have a for loop where you call imcrop with the bounding box, then call imwrite() to save the cropped image
for k = 1 : length(props)
    thisBB = props(k).BoundingBox;
    croppedImage = imcrop(rgbImage, thisBB);
    filename = sprintf('Image #%d.png', k);
    imwrite(croppedImage, filename);
end

Let me know if you have any trouble.

[Emre Havan]

Thanks for the answer, this approach makes sense but when I try to run it with noised image, it captures 4 dots around, the outer square, and some noises as squares.

Then I try it with somewhat denoised image but then it only captures some areas of black dots, even if I try to capture 32 images, it stills keep capturing around the big dots around the corners.

[Stephen Otieno]

@imageanalyst

I have a similar image, the only difference is that there are no black borders. I am trying to find the colour of each square using Lab or Hsv and returing a array of the colour loactions in the image.

[Image Analyst]

Stephen: pretty easy. Start a new question and define EXACTLY what you mean by "colour loactions". Do you want the centroid of the squares and circles? Do you just want, for every pixel in the image, a list of [x, y, r, g, b] in a CSV file? If so, see attached demo. If not, start a new question and I'll answer there.

[Stephen Otieno]

ImageAnalyst: I think i used the wrong description by saying "colour location", what i want is the centroid of only squares and I want to select a group of pixels in each sqaure to determine the colour and getting the values in a matrix of 4x4 in a cvs. Only using either lab or hsv.

thanks