Extract heatmap on top of image

Question

I have an image like this one and the corresponding background image.

I am trying to extract the heatmap of the first image. My first approach was to mask every pixel that was not "equal enough" (thresholding the euclidean distance between every pixel), but the results were not good enough

Looks like a straightforward problem in image processing, but I lack experience. Thanks!

Answer 1

Obtaining the exact heat map is a highly non-trivial problem since the heat map was superimposed with varying levels of transparency.
This means that per pixel there are two unknowns:

1. The heat map itself and
2. transparency (they are linked but the exact mapping is unknown).

If a rough segmentation suffices, I suggest a simple thresholding via the color intensities. You can see that the stronger regions of the heatmap are surrounded by a strong, blue color:

I = imread('https://i.stack.imgur.com/7oVZK.png');

% Simple, manual color thresholding
mask1 = I(:,:,3) > 230 & I(:,:,1) < 30 & I(:,:,2) < 10;

Now we dilate the mask to close any holes in the blue heat map boundaries and fill the holes:

% Morphological processing
mask2 = imdilate(mask1, strel('disk',5,0));
mask2 = imfill(mask2, 'holes');

Finally, we can extract the heat map:

% Extract heat map
heat_map = I;
heat_map(~repmat(mask2,[1 1 3])) = 255;
imshow(heat_map)

Here is the Result

Edit: If you have to process many similar images, it might be more robust to perform the segmentation in hsv space:

I_hsv = rgb2hsv(I);
mask1 = I_hsv(:,:,1) > 0.6 & I_hsv(:,:,1) < 0.7 & I_hsv(:,:,2) > 0.95 & I_hsv(:,:,3) > 0.9;