Info: The Matrix - with original theatrical color timing? — Page 6

Maybe my memory is shot, but I could have sworn that the BD and DVD greens were a different shade.  The BD being more yellow/lime.

I'm not seeing that in the comparisons here.

Doctor M said:

Maybe my memory is shot, but I could have sworn that the BD and DVD greens were a different shade.  The BD being more yellow/lime.

I'm not seeing that in the comparisons here.

The BD changed it to the much brighter more pronounced shade of green used in the sequels. The DVD was always much more yellow in certain scenes, though it does vary.

EDIT: Forgot to say, those shots posted by You_Too look amazing. That's exactly how I'd like The Matrix BD to look. :)

You_Too said:

I can't help but think you must have done something wrong when you tried ColourLike. (Maybe you didn't run it in RGB32?) 

Ehr... should I run ColourLike in RGB32? As it is compatible with YV12, and I prefer to avoid color space conversion, I didn't... well, as you proved it could be done with very good results, I'm pretty sure it could be possible to use it for the whole movie, but I have not time now to test it - I'll do it on Sunday and, as there are two movements (blue pill and red pill) I thought to do the restoration using the ColourLike (or the best alternative) to match colors as close as possible to the DVD, and make a fanedit using ChromaRite; in the future, if it turns out that the DVD is not valid as color reference, we'll see...

I've done a fast test with ColourLikeRGB, making the histograms every 1000 frames (last time I used it on YV12, I made every one frame).

Well, I don't know if it's because I wrote the histograms every 1000 frames, or if I've done something wrong with the script, or with my PC... but here you are the result - not really different from the ColourLike I've used until now:

and here you are my script:

bd .ConvertToRGB32().WriteHistogram("histograms\bdRGB.txt",1000)

dvd.ConvertToRGB32().WriteHistogram("histograms\dvdRGB.txt",1000)

CLRGB=bd.converttoRGB32.ColourLike("histograms\bdRGB.txt", "histograms\dvdRGB.txt").converttoyv12

stackvertical(\
bd.subtitle("BD",size=60),\
dvd.subtitle("DVD",size=60),\
CLRGB.subtitle("ColourLikeRGB",size=60)\
).spline36resize(960,1200)

You_Too, can you help me, please?

Andrea: Colourlike doesn't make a new histogram every 1000 frames when you set every=1000. It samples the movie in 1000-frame-steps and still makes one histogram. All the every argument does is change the sample size.

My script generates a new histogram for every frame. That's the difference.

You should see results immediately with my script because it's not doing a WriteHistogram for the entire movie at one time, just the frame it's currently on.

You know what? Keeping posting in threads like this is always "food for thought", and I learn (or correct) techniques, methods or scripts every time - not counting the constructive comments, everytime well accepted. Thanks to all the forum members!

You_Too, thanks for your help - AntcuFaalb, thanks for your script.

Now it seems to work, and the results are the same of the image You_Too posted. But, as I have the DVD, I could test every scene I'd like to... so, what about these?

While the first result is bad, the second is awful and unusable... What do you all think I should do for these scenes? It will be better to use the BD, the upscaled DVD (I've done it for the second scene, check out the last test clip some posts before - or (the best option to me) - to see if the DVD chroma lies perfectly (or almost) onto the BD luma?

Andrea: Try g-force's script instead: http://originaltrilogy.com/forum/topic.cfm/New-colormatching-script/topic/15002/

Tested. Good results, but not always.

Still got problems with your script, every now and then, don't know why...

I changed my ColourMatch script yesterday; now that I tested your script, ColourMatch and ColourLikeFBD results are the same (in good and bad scenes...) but it actually works in every frame in my PC, so I think I'll stick with it - I'll post some screenshots soon.

Wow. Those shots really show that the problems with the BD go far beyond just the green tint!

Really great to see the progress being made. Thanks to all involved.

You_Too said:

Yeah, unfortunately, ColourLike seems to have troubles handling scenes with crushed blacks or whites.

Not only ColourLike... my script called ColourMatch, despite the name, is not using ColourLike or g.force RGBMatch, but behaves in the same manner... RGBMatch at the contrary have not problems with the second scene, limiting to imitate the BD colors, but is not very good in matching colors...

So, I definitely use the DVD upscale for the second scene (very few seconds, maybe three or four if not less) - again, download the test clip and check.

I'll check if DVD chroma lay (almost) perfectly on the BD chroma for the first scene - too long to fool a trained eye of our forum members... (^^,)

Time to sleep!

_,,,^..^,,,_ said:

I've done a fast test with ColourLikeRGB, making the histograms every 1000 frames (last time I used it on YV12, I made every one frame).

Well, I don't know if it's because I wrote the histograms every 1000 frames, or if I've done something wrong with the script, or with my PC... but here you are the result - not really different from the ColourLike I've used until now:

This shows off the particulars of the DVD's color, and what is presumably the 35mm/method of shooting. The BD opts for contrast boosts and slathering everything in that fake green. This color correction fixes this but results in an image that is rather flat. The DVD has a uniform greenish-yellow cast to all the in-Matrix skintones, and a unique look to everything else in and out of the Matrix. This is lost on the SEDVD/BD and isn't visible in the attempted color rendering.

I'm thinking that certain filters were used to influence the film as shot, depending on whether the scene was in and out of the Matrix--and then timed deliberately further in post to create an effect similar and IMO far more effective than the green of the sequels.

35mm cells would really help, but the VHS looks the same as well.

First thing: I realized that setmemorymax and setmtmode avysinth commands made strange effects on my previous test, so I deleted them and now results seems perfect - no problem with ColourLikeFBF anymore.

After I finished the color correction tests for "The Thing", I thought to use the most successful technique with "The Matrix"... what I have discovered is that if the (upscaled) DVD chroma is applied over the BD luma, we have the best compromise (IMHO). But, as upscaled DVD has almost never the same frame crop of the BD, chroma "bleeds" and we have a bad result.

Mission: try to match as close as possible the BD luma + DVD chroma without overlay the actual DVD chroma onto the BD luma... and the mission seems accomplished overlaying the ColourMatch chroma onto the BD luma.

Essentialy, ColourMatch do the same thing ColourLike (and ColourLikeFBF) do - mimic the color of another video - but ofter results are more accurate; here you are some comparison screenshots, including ColourLike and ColourLikeFBF for reference.

WARNING! when you read "ColourLikeRGB", I'm referring to ColourLikeFBF (histogram written for each frame) applied to converted RGB video - sorry AntcuuFalb, too lazy to reupload all the screenshots! (^^,) - while ColourLike is used "as is" without converting video to RGB (one histogram for the whole video, taken every 1 frame).

***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   *** 

CL is really different, and CLRGB is in between CL and DVD, while BD+CM is almost perfect.

***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***

CLRGB matches very well the DVD, and BD+CLRGB is close to BD+CM.

***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***

CL and CLRGB are similar between them, but both different from DVD, while BD+CM is almost perfect.

***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***

This is one of the most irritating scene to me: "Neo with green beard" on BD... as you can see, both CL and CLRGB show "green beard",

BD+DVD colors are quite good, but as DVD cropping is different, Neo seems a bit "zombie"... BD+CM seems the best.

***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***

This time CL is more faithful to DVD, and in this scene DVD, BD+DVD and BD+CM are really, really close to each other.

***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***   ***

Conclusion: I think I should use this technique, and proclaim result as restoration and not fanedit... infact, I'm not doing color correction "by eye" matching a third color reference source, or making a personal color timing scene-by-scene, but only using BD as luma reference and DVD as chroma reference; this preservation could be considered the "son" of BD and DVD, and you know that sometimes, child takes the best from his mother and his father...

Opinions?

The problem with the BD luma seems to be that the contrast is too high as illustrated in the third shot with the darkening of Neo's jacket and over-whiteness of the background. Would there be any way to correct or compensate for this?

In many of the shots ColourLikeRGB looks closest to the DVD reference to me but, as you say, those green beard/head shots really stand out.

rockin said:

BD luma, DVD chroma looks the most overall balanced to me.

I agree with you, but as I explained, it's not possible to use this technique, as the DVD and BD cropping are different (take a look at the "Neeo with green beard" screenshot...); as DVD luma + ColourMatch chroma is, maybe, 95% or better close to it, without side effects, this it the reason I decided to use this last method.

The only thing that it needs fixing is the contrast.

I could set the contrast lower to match better the DVD, but what if the DVD contrast was wrong?

We need more input from other people.

I agree.