Harmy, by now you know how anal I am about this stuff, so about color let me say that I will absolutely be offering "final/this-is-really-got-to-be-what-it's-supposed-to-be" images, but I'm just not ready to call anything final yet. It won't be forever. In fact, probably within the next couple of months I'll have some things together which I'd feel comfortable taking to the bank.
It's a cluster. Let's say, theoretically, you had multiple near-pristine Tech prints, and you noticed almost zero, or negligible, color variations between them. You might be prepared to sum the prints, and take the average value for a given pixel and say "that's it," except the nature of film is such that one frame later, it's different. So which one is it supposed to be? Well, let's say you then took the sampling of every occurrence of a particular color (say, a costume) across the entirety of the film, and took the average. Well, that negates the deliberate color-grading which was done scene-to-scene, which intended for the color to appear different each time. So then perhaps you just take it scene-by-scene, and pick an average value within that scene, and declare a particular color only valid for a certain range.
That's closer to definitive. But that has a lot of room for error in it as well, such as drifts along the production chain, from negative to IP to IN to Print. That process is not 100% stable. And then, if they're prints, then they're old, and there's color fading to consider. And then, what about at the scanning stage. Are you sure you're capturing using a light temperature appropriate for the stock? I only know of one scanner even capable of dialing in the color temperature of its light source on a per stock basis! And then, are you capturing to a colorspace which can faithfully capture the range on the stock?
And then, what about your working/monitor environment?
To really do this right you'd need a slew of things:
1) Reference photos of original costumes, sets, and models taken in 1976 (common; can be found in the private collections of original vfx artists and LFL Archives), and ideally with color charts in them (extremely rare, but exist somewhat). If you have charts, you have, at least somewhere, a record of what each color REALLY was; I mean, in life, irrespective of how they ultimately lit and graded the images.
2) Data on the various stocks used when shooting Star Wars to determine their inherent biases and/or limitations in capturing color (this is obtainable). From this you at least know something about what was even possible to capture, compared to the actual color in life.
3) Data on the various stocks used along the transfer process on the way to print to determine their inherent response characteristics (this is obtainable), and response characteristics of your prints (obtainable).
4) Prints which haven't faded, or have barely faded. A pristine, or near-pristine Tech is about the best you could ever hope for on this front.
5) A scanner which tunes its light source to match the biases of the stock. Currently the only one I know of is DFT-Film's Scanity. Negative stock, for example, is orange tinted, so it uses a slightly green light source when capturing, the exact temperature of which is dialed in based on the stock.
6) A calibrated working environment, and flexible colorspace.
So, in the end, it's do-able. It's just an odyssey of work, research, and asset acquisition. Oh yeah, and a shit-ton of money. :)
When you consider all this, the fact that any of our fan-based corrections look as good as they do to as many people as they do, is pretty impressive, and something to be grateful for.
_Mike
