The color matching tool v1.1 has been released. The algorithm has been upgraded, such that creating a color matching model only takes a few seconds, even for a set of 4K frames. Please send me a PM, if you’re interested.
An updated version of the color matching tool is almost ready. It includes an updated version of the algorithm, that produces results in a matter of seconds even for a 4K image.
It looks amazing NeverarGreat!
Here’s the color balanced version of that still. You can clearly see the Death Star walls have a blue bias, as Mike Verta maintains should be the case:
Thanks y’all!
Here’s the shot of Luke regraded, based on the 1997 telecine (bluray top, regrade bottom):
I don’t like the 1997 telecine grading, too blue.
The 1997 telecine samples have not been corrected yet. The blue tone is removed in the second pass, like for the last examples.
Here’s the first duo of regrades, based on the 1997 telecine (top bluray, bottom regrade):
Here’s an example of how one of the above frames of the 1997 telecine would look after a second pass correction.
Before:
After:
Hey DrDre
Since you appear to be our resident “build algorithms from scratch” guy, I have a request.
Multiple projects on this site use some form of an average of multiple captures or sources.
I’ve seen a few different ways people are doing this, but they appear to be limited to mean and median averages. IMO, a mode average would be the preferred method of averaging in this specific instance, but I’ve never found a plugin or tool that implements mode. Avisynth would be the preferred platform, but perhaps it’s something you could implement in a matlab environment?
Is this something you could/would be willing to look into?
Not sure if I will have the time available, but I will dwell on it for a little while…
DrDre - you sir are a master. That is it. Perfection.
Thanks, I like them a lot also!
Good idea for another source to use! but… Some look a bit blue, some too bright, some with a lack of contrast. Not loving them I gotta say.
Do the same for the JSC, and the LPP, and the IB, and this one, throw out the outliers, average the results, and then you may have something.
-G
There’s still some color variation, because I applied a single correction to the entire sequence. The variation is part of the source sadly. A bluray matched to this source definitely would require a second pass to unify the colors, but many print scans are really no different. However, I think there’s something lurking in there…
DrDre - you sir are a master. That is it. Perfection.
Thanks, I like them a lot also!
I’ve been experimenting a bit with color balancing the telecines of the 1997 SE, to see what colors come out of this process. Here are some results:
/tons of screenshots with ultra natural skin tones
SOLD. Can those shots be used to correct the blu ray to those colors then?
That’s the plan… 😃
I’ve been experimenting a bit with color balancing the telecines of the 1997 SE, to see what colors come out of this process. Here are some results:
Deleted
I was wondering if a copy of the JSC laserdisc is around usenet or spleen. Is it even digitally available?
Yes, Althor1138 made an excellent transfer of the trilogy, and it can be found on myspleen as a single bluray.
Rather than posting a bunch of random frames, like before, I will start with the Luke meets Obi-Wan scene, do the regrade for the entire scene, and present the results for all of you to critique. 😃
Now that poita and williarob are testing the color restoration tool, it is time to start working on the bluray regrade again. Updates will follow soon…
Wish I had more to contribute, but I’m afraid most of my comments here are just similar “you’re an amazing wizard, DrDre” comments. Keep it up 😃
Thanks! 😃
Poita and Williarob both have a beta version of the color restoration tool I’ve developed. Hopefully they’ll be able to put it to good use. Here’s an example of what the tool can do to aid the process of restoring the colors of the red faded TESB print scan, that poita will be using for his restoration:
That final correction of the emperor shot above is obviously a marked improvement over the untouched or the “non blue” versions, but it still feels a bit too pinkish.
This is why I said, you should in practise use a frame for calibration, that’s closer to the frame you want to correct. 😉 This was only an example to show the process, not an example of a final correction.
The color balancing/restoration algorithm not only restores color, but also restores detail, as the detail is in the color information. Here’s an example for the severely red faded 70mm print of The Empire Strikes Back:
Here’s a screenshot comparison:
A thought just occurred to me. Wouldn’t it be useful to have the border scanned as well? My understanding is that it is printed using the photographic process, so it is the 3 color black and should faded in the same way. So couldn’t it be used to better predict the original unfaded colors? Just a thought. I like the way your algorithm works. It makes some very nice images.
Yes, that would certainly be very helpful, if it is available.
Exactly, take the example of the red shifted ROTJ frames.
Correcting both frames indidually removes the red cast:
However, for the second frame the delibirate blue cast is also removed. Assuming the red cast is similar for the entire film, I can build a color restoration model based on the first frame, and use it to correct the second frame:
Now, the red cast is removed, but the delibirate blue cast is retained. In practise it is better to use a frame closer to the second frame for constructing the color restoration model, but you get the idea.
The color correction tool v2.2 has been dubbed the color matching tool v1.0 to distinguish it from the upcoming color balancing tool v1.0. The functionality of the color matching tool v1.0 is the same as the color correction tool v2.2.