Color matching and prediction: color correction tool v1.3 released! — Page 10

However, if you have enough references available, the multi color space match/predictions within a shot are the way to go:

Bluray:

Bluray matched to Tech IB with single color space method:

Bluray matched to Tech IB with multi color space method

It has to be said that the Star Wars bluray is notoriously difficult to regrade. In other cases the single color space method works really well, like for the Raiders 35 mm frame, and as such it's predictions are more reliable:

35 mm:

Bluray matched to 35 mm with single color space method:

DrDre said:

Here are the predictions for some of the other frames:

So, in principle you could regrade the entire film, based on the single frame with this method, and it would look pretty awesome. I'm sure these predictions much closer to the 35 mm colors than either the bluray or WOWOW. Since this method is much faster, it would take about 2.5 days to do the entire film. Oh, and those nazi flags are definitely red in this regrade.

 If one were to regrade the entire film, what would be the best way to do it? Export the whole film into images via e.g. virtualdub and load them using your program? It would certainly be awesome to watch Raiders with those colours.

Kurosawa10 said:

DrDre said:

Here are the predictions for some of the other frames:

So, in principle you could regrade the entire film, based on the single frame with this method, and it would look pretty awesome. I'm sure these predictions much closer to the 35 mm colors than either the bluray or WOWOW. Since this method is much faster, it would take about 2.5 days to do the entire film. Oh, and those nazi flags are definitely red in this regrade.

 If one were to regrade the entire film, what would be the best way to do it? Export the whole film into images via e.g. virtualdub and load them using your program? It would certainly be awesome to watch Raiders with those colours.

 Since it won't take much time, I'm considering doing this regrade myself. I'll start a new thread on the subject.

DrDre said:

Kurosawa10 said:

DrDre said:

Here are the predictions for some of the other frames:

So, in principle you could regrade the entire film, based on the single frame with this method, and it would look pretty awesome. I'm sure these predictions much closer to the 35 mm colors than either the bluray or WOWOW. Since this method is much faster, it would take about 2.5 days to do the entire film. Oh, and those nazi flags are definitely red in this regrade.

 If one were to regrade the entire film, what would be the best way to do it? Export the whole film into images via e.g. virtualdub and load them using your program? It would certainly be awesome to watch Raiders with those colours.

 Since it won't take much time, I'm considering doing this regrade myself. I'll start a new thread on the subject.

 Awesome!

Someone asked this before and I'm not sure if there was an answer. But, if going the route of a single correction applied to an entire film, is there a way to simply make an LUT out of the original and modified frame? 

towne32 said:

Someone asked this before and I'm not sure if there was an answer. But, if going the route of a single correction applied to an entire film, is there a way to simply make an LUT out of the original and modified frame? 

 In essence, that's the output of the algorithm (for the single color space model). For 8 bit and 16 bit images this can be done, exactly. However, for the exr images this is not possible, as there are an infinite number of possible color combinations, which is why it is a HDR image format. 

In the Raiders color timing thread it was remarked that the bluray was an attempt to approach the theatrical color grading, and that the original color grading was more yellow, especially in the desert scenes, as is evident from this 35 mm frame:

Well the color correction model predicts the following for desert shots:

Bluray:

Bluray corrected with single frame color correction model:

I think the color correction model nailed it.

If you have some $$$$ you can use this to make a LUT:

http://www.lightillusion.com/lut_image.html

There is probably a cheaper way to do it though.

I believe I have solved the problem of crushed whites, and blacks for the CollorCorrect GUI v1.2. While trying to improve the grading for a frame of The Empire Strikes Back with a color correction model I calibrated on a frame of Star Wars, I had the following problem:

Semi-Specialised Edition V2.2:

This is what I got:

The source frame suffers from crushed whites, so the model could not correctly estimate the colors, resulting in artifacts.

I realized where the problem is with crushed whites and blacks, namely that it introduces errors in the color distribution, implemented a solution, and this is what I got:

Pretty neat, I think, and the colors look pretty good as well... :-)

Do you think you can make the algorithm avoid crushed whites on the result, even if the source frame has them?

If you make a LUT and download DaVinci Resolve you can grade in real-time through the LUT.  Placing a grade before the LUT will allow you to save clipped blacks and highlights.  Then you can then render an entire film in real-time.

Laserschwert said:

Do you think you can make the algorithm avoid crushed whites on the result, even if the source frame has them?

This is exactly what it does, it corrects for crushed blacks and whites in both the source and the reference, to a certain extend.

Don't 35mm prints have crushed blacks and whites too? It's not the kind of "posterized" crushing of digital, but to provide one example, I recently saw an LPP print of The Goonies where those overcast/rainy Astoria skies looked almost white at times. Though there was still visible cloud detail, those skies were noticeably blown out white-gray.

With the limited color space we're working with, isn't it impossible to completely avoid crush/blowout/detail loss? Again, 35mm prints had less detail than the higher-generation sources these transfers are coming from, but the characteristics of photochemical timing/printing, print grain, and theatrical projection make it less obvious and less blatant and more "natural" than in the digital realm.

nightstalkerpoet said:

So just to clarify,  the single frame technique corrects all resulting frames in a single color space? 

If so, is there a chance to implement a third option, where the original image is corrected, saved, and then the multiple color space model analyzes the source image and uses the new image as the reference? I'm just thinking that some images in the film may require a little more force to push the correction. Or is that assumption incorrect?

 That would not work, because the multiple color space method would essentially immediately converge to the single color space method, because the initial correction effectively removes all the color information in the other spaces.

CatBus said:

DrDre said:

I believe I have solved the problem of crushed whites, and blacks for the CollorCorrect GUI v1.2.

I have to say, this might change everything for me.  Back when you were color correcting the Blu-ray exclusively, I was inclined to believe that contrast boosting on the Blu was destroying your highlight detail.  But then I saw the same thing when you corrected the GOUT and I thought something must be wrong with the method in general.  I didn't complain because frankly I felt you were catching quite enough flak already, and were still doing interesting work in spite of the problems.

Now... things look very promising indeed.  Keep up the good work, and I'd like to see more captures using this new method.

 Could you explain with an example? Contrast boosting generally results in crushed highlights. There's pretty much nothing you can do about this. The detail in de mid range is enhanced, at the cost of the highlights. 

CatBus said:

DrDre said:

CatBus said:

DrDre said:

I believe I have solved the problem of crushed whites, and blacks for the CollorCorrect GUI v1.2.

I have to say, this might change everything for me.  Back when you were color correcting the Blu-ray exclusively, I was inclined to believe that contrast boosting on the Blu was destroying your highlight detail.  But then I saw the same thing when you corrected the GOUT and I thought something must be wrong with the method in general.  I didn't complain because frankly I felt you were catching quite enough flak already, and were still doing interesting work in spite of the problems.

Now... things look very promising indeed.  Keep up the good work, and I'd like to see more captures using this new method.

 Could you explain with an example?

The stormtrooper on the far right's chest:
http://screenshotcomparison.com/comparison/139124

Artifacts in the highlights on C-3PO's leg:
http://screenshotcomparison.com/comparison/140781

Lots of highlight detail lost here:
http://screenshotcomparison.com/comparison/142085

Seeing those shots redone with the new method would be very interesting to me.

 The first two were actually done with a different model, than the one I've been using for a while now. The last example is actually due to lack of color depth and noise in the GOUT. I will try to show some more examples, but the new method was used since page 6 of this thread. 

nightstalkerpoet said:

Which color space are you working in for single match mode? I'm still getting (imo) better results with my gimp process - take a look at my steps and see if there's anything about it that can be implemented in the gui.

http://screenshotcomparison.com/comparison/143328

 You should try your method on the skeleton frame. I've not had good experience working with YCbCr. When I use it, the blue in the background is muted.

Also in this frame, the sky is more green, and the wires on the right have virtually disappeared.

Can the falloff on the blacks and the highlights be less linear?  I typically isolate highlights and try and recover them as much as possible during grading.  As long as it looks natural.

DrDre said:

I believe I have solved the problem of crushed whites, and blacks for the CollorCorrect GUI v1.2. While trying to improve the grading for a frame of The Empire Strikes Back with a color correction model I calibrated on a frame of Star Wars, I had the following problem:

Semi-Specialised Edition V2.2:

This is what I got:

The source frame suffers from crushed whites, so the model could not correctly estimate the colors, resulting in artifacts.

I realized where the problem is with crushed whites and blacks, namely that it introduces errors in the color distribution, implemented a solution, and this is what I got:

Pretty neat, I think, and the colors look pretty good as well... :-)

 Turns out the solution for crushed whites and blacks reduces the accuracy of the match. So, back to the drawing board for this problem.

nightstalkerpoet said:

That's what I'm getting with prediction based on the alley shot.

 That's what I got as well, when working in Ycbcr color space. In RGB you get this: