Hi all,
I'm trying to remove known backgrounds from some images, which is easy enough in itself -- except some of the images and backgrounds are overlaid with random "noise." I'm wondering what's a good way to get a "good enough" result.
The basic setup is that I have two images (screenshots from a game). The first image is the background. The second image is the background plus an overlay image. I'm after the overlay image. So I insert the second image as a layer over the top of the first image and set the new layer's blend mode to "Difference". Flatten and voila! Works a treat.
Trouble is, there's additional, semi-random information overlaid on both the first and second images in some cases (eg: a rain animation).
So, I take multiple screenshots of both images. Each will have a different noise pattern. I'd like to combine all the versions of image 1 into a base background, and all the versions of image 2 into a base overlay, but I can't figure out a good way to do it.
Ideally, I figured to use a "Majority Report" approach; that is, for each pixel, take the value that appears in the majority of screenshots. Or, assuming there's no identical match, take the average of the closest two values. I definitely can't see any way to do this within Paint.NET.
Alternatively, I figured to average the three screenshots, but I can't see that feature, either.
So now I'm wondering, is there a plugin that will do either of these things?
Or if not, is there another way to get much the same result. Assuming the background image has three versions (1a, 1b, 1c) and the overlay image has three versions (2a, 2b, 2c), maybe I could do some clever maths.
i) a Difference a Difference b (from top to bottom) inverts the brightness of the noise differences -- not really a solution.
ii) 2a+2b+2c Difference 1a+1b+1c might result in some effective averaging, but the additions could easily clip.
iii) Something tricky with setting the opacity...
OK, so I'm clearly not a pro at image manipulation, and certainly not blending algebra. But I'd love to learn!
What do people suggest?
(Big thanks in advance.)