Hey, Posh -
Let me re-state in mathematical terms what I believe is your question:
Regard all the processing that was done on an image as a mathematical function that takes each triple of three variables, say, (H,S,L), and maps them into different values (H', S', L'). If I understand you correctly, based on a large number of pixels in one before-after image pair, you want to come up with a good estimate of that rule so that you can apply it to other things added to that image, apply it to entirely different images, etc.
If that's what you want, that's a very, very interesting question. However, as far as I know, there is no feature built into PS, nor in any add-ons / plugins that I am familiar with that will do what you ask in anything approaching an automatic manner.
This is not to say that it couldn't be done, just that if it has, it's probably only available in the technical image processing literature. As I get some time, I'll keep this in mind and see what I can find, but it probably will be an algorithm written using the Matlab Image Processing Toolbox, or some another programming language, so it probably wouldn't be of that much use to you.
However, let me offer a possible DIY workaround:
1. Put your "before" image at the bottom of a layer stack. (Layer 0)
2. Put your "after" image above it. (Layer 1)
3. Change the blending mode of L1 to either "difference" or "subtract".
4. In preparation for needing to amplify small differences, put a "Levels" adjustment layer at the very top of the stack. (Layer 2). To amplify small differences, move the two endpoint sliders in towards the center.
5. Here is where the art of the method and and experience comes in. Start introducing adjustment layers between L0 and L1 as you see fit to try to make L0 + the effects of the adjustment layers look like L1. You can do this initially by eye, but eventually, you will get to the point where you want to magnify small differences. This is where L2 comes in.
6. When you get a set of adjustments that transform L0 into a good approximation of L1, you can now open another image and apply the exact same set of adjustment layers (together with all of their settings) over to the new image. This certainly isn't automatic, and depending on how good you are, it may be only a rough approximation of what you need, but it's better than nothing.
There was a thread on exactly this method maybe 6 - 9 months ago. It included an example. I'll try to find it for you.
Great question.
Cheers,
Tom
