You_Too said:
If you haven't seen this post by Harmy ...
Thanks! Not only that, but the whole page is great -- and now, of course, I must go through the entire thread. :) Actually, that page demonstrates the virtues of working in RGB in a number of ways.
Harmy's previous post, there, showed his quick correction (using Affect Effects?) to match another source. Here's the original scan with luminance graphed (split apart RGB's are roughly the same, although 2 channels had slightly blown-out whites):
It's in pretty good condition.
However, the adjustment compressed the entire range and produced slightly crushed blacks:
While "the color" part may look better (from the program interpreting the dials and sliders), the frame is more damaged overall (which would have been immediately noticeable working on the RGB elements directly).
Another instance of RGB superiority is in noise reduction. Another scan from 70mm Star Wars Film Cells (http://www.jedi1.net/) is very noisy, but split-apart RGBs show where most of that noise resides (here cropped but full sized):
(From what I've seen, this distribution of noise seems to be the norm. I wonder if it's from the very construction of film itself?)
By applying varied noise reduction (temporal would have been superior, but only rudimentary spacial is available in the paint program), it isn't applied where not needed (minimal loss of remaining detail). In my quick demonstration, RED is untouched, GREEN has a little smoothing to approximate RED's noise level, and BLUE has more to get closer to RED's without total smearing (which it seriously needs). A slight overall sharpening was added after the RGB rejoin, to offset the blunderbuss of this spacial smoothing:
