Apply mask to RGB image. If mask is RGB as well do i have to seperet the mask to its channels too?

Hello I want to apply this mask onto this RGB color to use in NMF

...

function [ R, B,G ] = Split_Image( )
 
Input=im2double(imread('dataset/transport_clean.png'));
mask=im2double(imread('dataset/transport_mask.png'));
Q=mask(:,:,1); 
R=Input(:,:,1);
G=Input(:,:,2);
B=Input(:,:,3);
 
X_red = Q.*R;
X_gre = Q.*G;
X_blu = Q.*B;

imshow(X_red);
end

So I want to inprint the left image on the right image to get an output in the third image.

How ever in my case i will have a red blue green channel with the masks.

I will then process these 3 images seperately using nmf and combine them after to reconstuct the original image

How ever when i run my code i get the same image below for all channels. What am i doing wrongly

Answer 1

[Subhadeep Koley]

Hi, refer the code below.

function [ R, B, G ] = Split_Image()
    input = im2double(imread('transport_clean.png'));
    figure; imshow(input, []); title('Input image');
    mask = imcomplement(im2double(rgb2gray(imread('transport_mask.png'))));
    added = input + mask *0.3;
    R = added(:, :, 1);
    G = added(:, :, 2);
    B = added(:, :, 3);
    figure; imshow(added, []); title('Mask over the input image');
end

[fadams18]

Hi thanks for your time.

Could you please tell me whats the differenct between using imcomplement

 mask = imcomplement(im2double(rgb2gray(imread('transport_mask.png'))));

and doing the code below.

mask  = double( mat2gray( im2double(imread('transport_mask.png')) ) == 1 );

[Subhadeep Koley]

Hi, if you use imcomplement, then only the portion of the mask which has the writing (text) will be added with the input image. As, the white portion of the orignal mask (where there is no text) will be "0" after imcomplement. And, input + (0 *0.3) = input.

In this way the mask will not modify the other portion of the input image.

[fadams18]

Last point please, Im confused on how u added the mask. you did

 Input + mask

Im used to

mask .* input

Whats the difference please.

[Subhadeep Koley]

Yes, you're used to multiply the mask with the input. However, in my code I have used a technique called Alpha Blending. I have added the mask with the input image where, the mask is modulated with a Alpha (0.3 in this case) value.

If you increase the Alpha value then, the text will be more prominent in the resulting image.

[fadams18]

Thanks a lot much appreciated.