Is there a way to extract the gray background from the image and just to keep the rainbow

I have the following image produced by a diffraction grating on chip producing diffraction pattern.

Is there a way to extract all the color spots while removing the noise of the background?

What I tried was clustering but I don't think is the tool for this task.

Here are all raw images: https://drive.google.com/drive/folders/1B3K4JZd0QXebFceb09hsIKSUl4zCdyzn?usp=sharing

% Read image
img = imread('rainbow.png');
img = im2double(img);

% Reshape image into N x 3 color pixels
pixels = reshape(img, [], 3);

% Number of color clusters (you can adjust this)
k = 5;

% Run k-means clustering on the colors
[idx, C] = kmeans(pixels, k, 'Distance', 'sqeuclidean', 'Replicates', 3);

% Reconstruct clustered image
clusteredImg = reshape(C(idx,:), size(img));

% Show original and clustered version
figure;
subplot(1,2,1);
imshow(img);
title('Original Image');

subplot(1,2,2);
imshow(clusteredImg);
title(sprintf('Color Clusters (k = %d)', k));

Answer 1

[Mathieu NOE]

hello

here a quick answer - as I understand you want to extract the rainbow area from the entire picture . You can do a boundary + inpolygon action as shown below :

% Read image
img = imread('rainbow.png');
imgray = rgb2gray(im2double(img));
[m,n] = size(imgray);

figure
imagesc(imgray)
axis square
colorbar('vert')

% image tresholding
ul = max(imgray,[],'all');
ind = find(imgray>0.6*ul);
[y,x] = ind2sub(size(imgray),ind);
% select rainbow boundary
k = boundary(x,y,0.75);
hold on 
plot(x(k),y(k),'r')

●

% keep only "rainbow" portion of image 
[X,Y] = meshgrid(1:n,1:m);
[in, on] = inpolygon(X,Y,x(k),y(k));
img_out = NaN(size(imgray));
img_out(in) = imgray(in);
figure
imagesc(img_out)
% colormap( [0 0 0; parula(256)] ); % if you want NaN be displayed in black
colormap( [1 1 1; parula(256)] ); % if you want NaN be displayed in white

axis square
colorbar('vert')

[Vahram Voskerchyan]

Yes, this is exactly what I wanted. Thank you!

[Mathieu NOE]

as always, my pleasure !