3 images fused into RGB placed upon one another

3 images into RED,GREEN and BLUE ,placed onto one another and with the background being black

cl_img_1 = init_img_gs('img1.png');
cl_img_2 = init_img_gs('img2.png');
cl_img_3 = init_img_gs('img3.png');

img_out= cat(3,cl_img_t_gs,cl_img_c_gs,cl_img_p_gs);
figure(4), subplot(1,1,1), imshow(img_out);

[Rik]

How do you decide which color should take precedence? There are 8 combinations, but you only want 4 colors (black+RGB), so you need to decide.

Answer 1

[Voss]

A = imread('triangle.png');
B = logical(imread('circle.png')); % convert uint8 to logical to be consistent with A and C
C = imread('parr.png');

AA = zeros(size(A),'uint8');
BB = zeros(size(A),'uint8');
CC = zeros(size(A),'uint8');

AA(A & ~B & ~C) = 255; % red is 255 where the triangle exists and the others don't
BB(B & ~C) = 255; % green is 255 where the circle exists and parallelogram does not
CC(C) = 255; % blue parallelogram on top
out = cat(3,AA,BB,CC);
imwrite(out,'finaloutput_test.png');

Answer 2

[DGM]

I can only assume that things have been edited, because as it stands, there isn't really a question. Since it's accepted, I'm going to assume that Voss's composition is the intended layer and color sequence.

How would I approach it? With MIMT tools as usual. Let's recreate the given composition.

% the images as read
A = imread('triangle.png'); % logical
B = imread('circle.png'); % uint8
C = imread('parr.png'); % logical

% layer colors (from BG to FG)
CT = [0 0 0; 1 0 0; 0 1 0; 0 0 1]; % any colors, any class

% compose the image one layer at a time
% replacepixels(FG,BG,mask)
outpict = replacepixels(CT(2,:),CT(1,:),A); % FG, BG are tuples, mask is I
outpict = replacepixels(CT(3,:),outpict,B); % FG is tuple, BG is RGB, mask is I
outpict = replacepixels(CT(4,:),outpict,C); % FG is tuple, BG is RGB, mask is I

imshow(outpict)

Note that in this composition, we have a mix of double, logical, and uint8 data, but the class mismatches don't matter so long as everything is properly-scaled for its respective class. The output class is inherited from BG, which in this case is CT(1,:) (i.e. 'double'). Also note that our inputs don't need to have the same size, so long as they can be unambiguously expanded to a common size.

Besides being class-agnostic, this workflow seamlessly supports both binarized and graduated masks. It also allows us to simply use any color sequence we want without being constrained to simple primary-secondary schemes, as happens with simple channel combinations.

% the images as read
A = imread('triangle.png'); % logical
B = imread('circle.png'); % uint8
C = imread('parr.png'); % logical

% say they're not binarized masks
A = imgaussfilt(im2double(A),5);
B = imgaussfilt(B,3);
C = imgaussfilt(im2double(C),1);

% any layer colors we want
CT = [0 0 0.2; 0.3 0 0.5; 0.6 0 0.5; 0.8 0 0.5];

% the composition is exactly the same
outpict = replacepixels(CT(2,:),CT(1,:),A);
outpict = replacepixels(CT(3,:),outpict,B);
outpict = replacepixels(CT(4,:),outpict,C);

imshow(outpict)

While it's doubtful that OP ever needed that sort of flexibility, it's not clear what OP actually asked for. The only thing I can do is provide information to passers-by.