Okay, I’m sold. Doing the whole reel at once looks pretty awesome.
Those tweaked shots look way too bright though, and I don’t even have my monitor brightness very high.
Yeah, I agree. I’m better at algo building than manual tweaking 😉.
This looks amazing, but one thing that bugs me: I don’t understand how an estimation for the first movie could be used for ESB and ROTJ. I don’t expect they were timed exactly the same.
It isn’t. Each movie is corrected without prior information or other references. All the information you need is in the movie itself. All you need to do is select a set of appropriate calibration frames from the print reel you want to correct, and the algorithm will estimate the correct colors. TESB and ROTJ are corrected independently.
Hey DrDre - would you be able to use this to fix something like Fellowship of the Rings EE that has had a blanket teal shift applied?
Preferred Saga:
1/2: Hal9000
3: L8wrtr
4/5: Adywan
6-9: Hal9000
Maybe to a certain extend, but this method is specifically meant for faded prints, which tend to fade in a predictable manner for each reel, depending on the way it was stored. LOTR was digitally manipulated, and generally the person doing the manipulation does not do it in a predictable manner.
A few things have happened over the last couple of days. Firstly, Williarob was kind enough to allow me early access to the Silverscreen Edition Grindhouse, to conduct some further color experiments. Secondly, I updated the color correction algorithm. After catching up on the in and outs of film, and digital film corrections, I realized I was working under the wrong assumptions, namely that I should work in the YCM color space, when in fact it should be RGB. I could go into the theory of it all, but it suffices to say, that everything fell into place, and I was able to correct reel 1 of the Star Wars LPP with a single color correction model without green shifts or any other major anomalies. I also ran some first tests on reel 2. Here’s a gallery of frames from reel 1, before and after, and one frame from reel 2. Surprisingly the colors are in many ways closer to the GOUT, than the Technicolor prints. No manual adjustments were made to these corrections. These are simply the result of a scientific theory for color fading.
Reel 1:
Reel 2:
Silverscreen Edition Grindhouse? Is that still going ahead?
Ol’ George has the GOUT, I see.
A few things have happened over the last couple of days. Firstly, Williarob was kind enough to allow me early access to the Silverscreen Edition Grindhouse, to conduct some further color experiments. Secondly, I updated the color correction model, and after catching up on the in and outs of film, and digital film corrections, I realized I was working under the wrong assumptions, namely that I should work in the YCM color space, when in fact it should be RGB. I could go into the theory of it all, but it suffices to say, that everything fell into place, and I was able to correct reel 1 of the Star Wars LPP with a single color correction model without green shifts or any other major anomalies. I also ran some first tests on reel 2. Here’s a gallery of frames from reel 1, before and after, and one frame from reel 2. Surprisingly the colors are in many ways closer to the GOUT, than the Technicolor prints. No manual adjustments were made to these corrections. These are simply the result of a scientific theory for color fading.
You never give up, that is amazing dedication. I hope to hear the reason for rgb vs ycm at some point when you have it sorted enough to explain. I always find the process behind new understandings to be quite educational.
Looks pretty good. All the shots on the tantive look a bit too dark/saturated to my eye, but the colors themselves look good 😃
Looks pretty good. All the shots on the tantive look a bit too dark/saturated to my eye, but the colors themselves look good 😃
If I remember correctly the LPP is a dupe of another print. The contrast of a print increases with each generation (something about optical gamma), similar to a 16 mm print, which is also generally a copy of a 35 mm print, so the scan should be gamma corrected for a better viewing experience, but it’s likely the original print was pretty contrasty.
I have found this rehearsal pic for the conversation of Han And Luke while they are going to the Big Sarlacc:
and it looked more natural than the ROTJ Grindhouse 35mm and Despecialized Edition 2.5 (since DeEd re-colourization almost matches a Technicolor Print, we all know that.)
so I adjusted an almost-the-same scene from DeEd 2.5 to match the photo. and slightly improved Luke’s shadow:
ROTJ DeEd 2.5 is not based on technicolor references because there were never any technicolor prints made of ROTJ
I have found this rehearsal pic for the conversation of Han And Luke while they are going to the Big Sarlacc:
and it looked more natural than the ROTJ Grindhouse 35mm and Despecialized Edition 2.5 (since DeEd re-colourization almost matches a Technicolor Print, we all know that.)
so I adjusted an almost-the-same scene from DeEd 2.5 to match the photo. and slightly improved Luke’s shadow:
That looks way better than any version I’ve seen!
Prequel Fan-Edit thread: http://originaltrilogy.com/topic/Yet-another-series-of-prequel-edits/id/17329
It’s not a useful indication of how the film actually looked though. It’s not about what’s “most natural” if it wasn’t “natural” to begin with.
It’s not a useful indication of how the film actually looked though. It’s not about what’s “most natural” if it wasn’t “natural” to begin with.
I agree. This thread is mostly about finding ways to rewind time, and bring the print colors back to it’s original state, warts and all. These do not necessarily have to be the most appealing colors.
The basic theory is pretty simple. Color fading happens as the dyes in the print degrade, which happens at different rates for the three different dyes (cyan, magenta, and yellow). The color distributions shift to brighter intensities. The distribution of the concentration of dyes is logarithmically related to the color intensity distributions of red, green, and blue. Over a large enough number of pixels/frames the average color is gray, which follows from the central limit theorem, and voila, you’ve got an algorithm.
For all of you interested in some theory on digital film correction, here’s a reference paper:
http://restorationfilms.free.fr/telechargements/CGIP2000.pdf
The color correction model in the above paper presupposes a set of desired colors. The color correction algorithm I developed in principle also works on single frames, under certain conditions, but does not use a set of desired colors like in the paper. So, here’s what my algo does for the example in the paper (low resolution):
Before:
After:
The algorithm naturally can also be used to correct faded photographs. I came across this website of a guy who specializes in restoring faded photos:
http://cyleow.blogspot.nl/2012/12/restore-and-recolor-my-world.html
As he states, it can take hours to manually correct a single photo or frame. This is what the algo does in a matter of seconds for two of his examples.
Before:
After:
Before:
After:
Another example of a faded photoghraph I found on google restored with a mouse click.
Before:
After:
Considering these results, I believe that the restored colors on the Star Wars LPP are probably very close to the original colors, and reinforce the idea that Star Wars during it’s theatrical run in 1977 looked a bit different from what Technicolor prints would have us believe.
So presumably the colours are identifiable from particles trapped in the interpositive?
Ol’ George has the GOUT, I see.
So presumably the colours are identifiable from particles trapped in the interpositive?
It’s all happening in the digital realm. It uses a model to describe the interactions between the various colors, and to model the chemical changes in the dyes on the photo or print.
Hm… I have a private 8mm film which is horribly red faded. I have tried various color correcting methods in the last 10 years but none of them worked as good as I hoped. Will you go public with the newest version of your tool?
Hm… I have a private 8mm film which is horribly red faded. I have tried various color correcting methods in the last 10 years but none of them worked as good as I hoped. Will you go public with the newest version of your tool?
That is the plan. I’m using the Star Wars LPP as a kind of test case, to discover the method’s strengths and weaknesses. The difficulty is coming up with a work flow that can be intuitively incorporated in a software tool. So, although it may take a little while, it will definitely be part of v3.0.
Here’s another example for an extremely yellow shifted photo.
Before:
After:
This is absolutely amazing. I’m dealing with faded Eastman prints sometimes when I’m dubbing old Eurospy flicks for CG and I’ve tried some free and commercial tools but none got ever near such impressive results. I can’t wait for the release to try it by myself. 😃
Okay… I must have this. I have a stack of about 200 photos my mother-in-law brought that my wife wants me to scan. most aren’t too bad, but some are almost pink.
So, it seems I might not be completely crazy for putting my trust in the GOUT as my guide for my color correction. I would very much like to see the full movie done with your color correction algorithm.
