How can i change the color of an image with another image?

I have two images. The first is the image that i want to show, and the other is the backgroud. I want to project an image that an operation with the output image and the backgroud image makes the input image. Any idea? Thanks!

[David Young]

I don't understand the question. Please could you give a simple example?

[Aitor]

For example, i have an image that i want to project, that we will call "imin". and i have the background where i want to project that image, that we will call "imback".

Then, the imout, the output image, will be a modification of imin, where the imback + imout = imin, and the projection of imout+imback is imin. I want to erase the background and project the imin in a not uniform background

I don't know if i explain myself. The imout will be a fusion of imback and imin, where imout + imback = imin

[David Young]

Sorry, but I still don't get it.

You define imin = imback+imout, but then you say that imin is the "projection" of imout+imback. What do you mean by projection?

If in fact imin = imback+imout, then imout = imin-imback. Your problem surely isn't that easy though?

You may need to give an example, by attaching images to your question. Or perhaps someone else understands better than I do.

[Aitor]

For example:

This is my input image (random image from internet that i'll like to project)

And this is a background

Its a non uniform background.

When i project the imin in that background, we will se the decoloration because it isn't a pure white background. So, i want to get the background, operate the input image and project this new image in the background. The combination between background and output (the one i want to generate) is the first image.

I don't know if i'm clear now.

[David Young]

Thank you for posting the images.

Sorry, but no, I still don't understand "project" or "combination", and I can't figure out what the output image should look like.

Maybe I'm failing to grasp something obvious. Does anyone else have more insight into this problem?

[David Young]

OK, Image Analyst has come to the rescue.

Answer 1

[Image Analyst]

If you want to do histogram matching, this is the best algorithm out there that I know of: http://www.eyemaginary.com/Portfolio/ColorHistogramWarp.html

Otherwise you may simply want to convert to hsv color space and just replace the h and s channels

hsv1 = rgb2hsv(rggImage1);
h1 = hsv1(:,:,1);
s1 = hsv1(:,:,2);
v1 = hsv1(:,:,3);
hsv2 = rgb2hsv(rggImage2);
h2 = hsv2(:,:,1);
s2 = hsv2(:,:,2);
v2 = hsv2(:,:,3);
newHSVImage = cat(3, h2, s2, v1);
newRGB = uint8(hsv2rgb(newHSVImage));

[Aitor]

Thank you very much! I'll try this and i'll comment if it works as i want! Thanks!

[David Young]

Wow! I'm impressed that you understood that this is about histogram matching!

[Thorsten]

In the example above rggImage1 is the gray background. You need not to transform it to HSV but can use it directly as the V.

First ensure that both images have the same size

I = imread('colorful-triangles.jpg');
B = imread('graybackground.jpg');
B = imresize(B, size(I,1)/size(B,1)); B = B(:, 1:size(I, 2));

Then create new RGB image by replacing the V plane of the color image with the gray scale image, as Image Analyst suggested

hsv = rgb2hsv(I);
hsv(:,:,3) = B;
newRGB = uint8(hsv2rgb(hsv));

An alternative approach would be to multiply the images

newRGBalt = im2double(I)*repmat(im2double(B), [1 1 3]);