DrDre

Hahaha! 

The results are extremely close to the actual Technicolor IB print references. 

This example of a sequence on Tatooine shows the power of the color transfer method I developed. First I calibrated the color transfer model on a frame from to bluray matched to the Technicolor IB print reference:

Next I used the color transfer model to correct the following sequence of frames from the bluray:

The results are...

I extended the method, where you match histograms in multiple color spaces. I did this in three ways:

1) Replace the histogram with an estimated probability distribution, using kernel fitting, which is easier to invert.

2) Invert the estimated cumulative probability distribution to obtain a direct mapping between the colors in the source frame and the reference in multiple color spaces. This can then be used to predict other frames.

3) Stabilize the solution.

Here is the script for SRV11:

orig=AviSource("Star Wars.avi")

orig=ConvertToRGB24(orig)

edi0=nnedi2_rpow2(orig,rfactor=2, cshift="spline64resize") 

edi=nnedi2_rpow2(orig,rfactor=2, cshift="spline64resize") 

sr1=SR(orig,1424,548)

sr1=ConvertToYUY2(sr1)

yuy2=ConvertToYUY2(orig)

sr2=SR(yuy2,1424,548)

rev=Reverse(orig)

sr1r=SR(rev,1424,548)

sr1r=Reverse(sr1r)

sr1r=ConvertToYUY2(sr1r)

yuy2rev=ConvertToYUY2(rev)

sr2r=SR(yuy2rev,1424,548)

sr2r=Reverse(sr2r)

s64=Spline64Resize(orig,1424,548)

s64=ConvertToYUY2(s64)

edi=ConvertToYUY2(edi0)

edi0=Spline64Resize(edi0,1280,544)

sr=Average(sr1,0.5,sr2,0.5,sr1r,0.5,sr2r,0.5,s64,-2,edi,1)

sr=Spline64Resize(sr,1280,544)

q1=QTGMC(edi,Preset="Placebo",Edimode="EEDI2",InputType=3,TR0=1,TR1=1,ProgSADMask=0,Blocksize=8)

q2=QTGMC(edi,Preset="Placebo", Edimode="EEDI2",TR0=1,TR1=1,Blocksize=8)

q2=SelectEven(q2)

Median(edi,q1,q2)

ConvertToRGB24()

edi=last

sr =Average(sr,1,edi,1,s64,-1)

edicd=ConvertToYUY2(edi0)

edicd=ConvertToRGB24(edicd)

Average(sr,1,edi0,1,edicd,-1)

nnedi2_rpow2(rfactor=2, cshift="spline64resize") 

Spline64Resize(1920,816)

I've done some tests on Empire, which is improved in terms of detail, but not as much as Star Wars.

@ Swazzy

The script can in principle be used for any SD material, although it has been written for videos with heavy aliasing. 

Here are three more frames from the bluray that have been matched to the Technicolor IB print:

This is a continuation of the color correction discussion in the super resolution thread...

zee944 said:

DrDre said:

confusedgambler said:

DrDre said:

For one reason only, because that's what the colors looked like in 1977. 

Come on, Darth Vader is solid black with no detail left. You can't seriously say that looks right or good or even "what it looked like in '77" Not a chance, no way.

Also,that's not what Mike Verta's version looks like and he has better sources to judge by.

Sorry to dissappoint you, but the reference is a scan of an unfaded Technicolor IB print, corrected to the print. Although there can be variation between prints, the colors are the way they are on the print. The reason detail is lost, is because the GOUT lacks the color depth to show the detail, but on the print it is definitely there. 

I agree with confusedgambler completely.

I'd like to say beforehand that I never trust a scan that I didn't do myself or at least I couldn't compare it with the print I held in my bare hand... I'm just skeptical. I've seen so many times so-called references turning out to be wrong.

But even if these are accurate copies of frames, when they're projected, they look a bit different depending on the projector/light bulb setup... And TV screen is also different. It would look horrible on TV if it looked like the corrected caps. Too deep blacks, burnt out highlights and possible oversaturation (maybe just due to heavy contrast). The untouched GOUT looks better, more natural and (ironically) more film-like. Sure it could use deeper blacks, a bit heavier contrast, but not that much.

If you're sure those scans are consistent with each other, then the colors could be used, but the luma levels are either should be disregarded or fixed by reducing the contrast. One setting for the whole footage. In my opinion, of course.

I also think the color matching script didn't handle this well, I don't think those artifacts particularly visible on the Obi-Wan comparison should be there.

The scans come from a reliable source, and look perfectly fine on HDTV. However, there are a couple of things to consider:

1) The GOUT's color depth is very limited, so although it is possible to match the colors, the color depth cannot be increased. This will result in artifacts. 

2) The GOUT is very noisy. As for the decreased color depth, the noise leads to artifacts, such as in the Obi-Wan frame. This has nothing to do with the algorithm. 

In short, crap in is crap out. 

Applying the corrections to the bluray leads to a much better result, with less artifacts, but even the bluray suffers from lack of color depth, digital noise, and crushed blacks:

Discussion about the color corrections can be continued in my color matching thread. 

As promised, here is the script for SRV12:

orig=AviSource("Star Wars.avi")

orig=ConvertToRGB24(orig)

edi0=nnedi2_rpow2(orig,rfactor=2, cshift="spline64resize") 

edi=nnedi2_rpow2(orig,rfactor=2, cshift="spline64resize") 

sr1=SR(orig,1424,548)

sr1=ConvertToYUY2(sr1)

yuy2=ConvertToYUY2(orig)

sr2=SR(yuy2,1424,548)

rev=Reverse(orig)

sr1r=SR(rev,1424,548)

sr1r=Reverse(sr1r)

sr1r=ConvertToYUY2(sr1r)

yuy2rev=ConvertToYUY2(rev)

sr2r=SR(yuy2rev,1424,548)

sr2r=Reverse(sr2r)

s64=Spline64Resize(orig,1424,548)

s64=ConvertToYUY2(s64)

edi=ConvertToYUY2(edi0)

edi0=Spline64Resize(edi0,1280,544)

sr=Average(sr1,0.5,sr2,0.5,sr1r,0.5,sr2r,0.5,s64,-2,edi,1)

sr=Spline64Resize(sr,1280,544)

q1=QTGMC(sr,Preset="Placebo",Edimode="EEDI2",InputType=3,TR0=1,TR1=1,ProgSADMask=0,Blocksize=8)

q2=QTGMC(sr,Preset="Placebo", Edimode="EEDI2",TR0=1,TR1=1,Blocksize=8)

q2=SelectEven(q2)

Median(sr,q1,q2)

ConvertToRGB24()

sr=last

edicd=ConvertToYUY2(edi0)

edicd=ConvertToRGB24(edicd)

Average(sr,1,edi0,1,edicd,-1)

nnedi2_rpow2(rfactor=2, cshift="spline64resize") 

Spline64Resize(1920,816)

zee944 said:

DrDre said:

confusedgambler said:

DrDre said:

For one reason only, because that's what the colors looked like in 1977. 

Come on, Darth Vader is solid black with no detail left. You can't seriously say that looks right or good or even "what it looked like in '77" Not a chance, no way.

Also,that's not what Mike Verta's version looks like and he has better sources to judge by.

Sorry to dissappoint you, but the reference is a scan of an unfaded Technicolor IB print, corrected to the print. Although there can be variation between prints, the colors are the way they are on the print. The reason detail is lost, is because the GOUT lacks the color depth to show the detail, but on the print it is definitely there. 

I agree with confusedgambler completely.

I'd like to say beforehand that I never trust a scan that I didn't do myself or at least I couldn't compare it with the print I held in my bare hand... I'm just skeptical. I've seen so many times so-called references turning out to be wrong.

But even if these are accurate copies of frames, when they're projected, they look a bit different depending on the projector/light bulb setup... And TV screen is also different. It would look horrible on TV if it looked like the corrected caps. Too deep blacks, burnt out highlights and possible oversaturation (maybe just due to heavy contrast). The untouched GOUT looks better, more natural and (ironically) more film-like. Sure it could use deeper blacks, a bit heavier contrast, but not that much.

If you're sure those scans are consistent with each other, then the colors could be used, but the luma levels are either should be disregarded or fixed by reducing the contrast. One setting for the whole footage. In my opinion, of course.

I also think the color matching script didn't handle this well, I don't think those artifacts particularly visible on the Obi-Wan comparison should be there.

The scans come from a reliable source, and look perfectly fine on HDTV. However, there are a couple of things to consider:

1) The GOUT's color depth is very limited, so although it is possible to match the colors, the color depth cannot be increased. This will result in artifacts. 

2) The GOUT is very noisy. As for the decreased color depth, the noise leads to artifacts, such as in the Obi-Wan frame. This has nothing to do with the algorithm. 

In short, crap in is crap out. 

Applying the corrections to the bluray leads to a much better result, with less artifacts, but even the bluray suffers from lack of color depth, digital noise, and crushed blacks:

confusedgambler said:

DrDre said:

For one reason only, because that's what the colors looked like in 1977. 

Come on, Darth Vader is solid black with no detail left. You can't seriously say that looks right or good or even "what it looked like in '77" Not a chance, no way.

Also,that's not what Mike Verta's version looks like and he has better sources to judge by.

Harmy actually addressed the differences between shadow detail on a print, and the GOUT:

http://originaltrilogy.com/forum/topic.cfm/Raiders-of-the-lost-ark-bluray-and-colour-timing-changes/post/670853/#TopicPost670853

Not surprisingly, Harmy's Despecialized Edition 2.5 more closely match the color references I have, than the GOUT in terms of shadow detail. In some ways the GOUT/Laserdiscs have really screwed up our perception of what details should or should not be enhanced for a theatrical version of Star Wars.

And here are the high color depth versions:

https://www.wetransfer.com/downloads/5be42c5bdc528d28dc05cce1cffbdb6e20150910095031/e840b10b02440de884b1f1468564ea9b20150910095031/04e505

Here's the same sequence for the Leia frame:

I'm not so sure there is less craft now than before. Working in the field of finance has thought me there's such a thing as a survivorship bias. We tend to forget the failures of the past, and remember the things that have lasted. In terms of filmmaking we remember the classics, but if I were to ask you what the worst films were of 1952, 1963 or 1977, I doubt anyone will remember. When we ask people about the great ones I'm sure most people could name a few. As we look further to the past, the ratio of good films over bad or mediocre films we remember tends to increase, giving us a false idea that there were relatively more great films during those days than there are now.

Here are the color corrected frames with full color depth:

https://www.wetransfer.com/downloads/bb04f4a775a8e503258f30b96e1a246720150910111126/ac0373632aeabae392346ec9f30e643920150910111126/02e62b

Thanks piota! Here's the frame corrected. The middle frame is corrected using Harmy's Despecialized 2.0 as a reference. The bottom has been corrected with a tweaked Harmy's Despecialized 2.0 to reduce the blue cast, and to have a more neutral color palette (for full resolution open image in new tab):

Here's a video sample for SRV14. I think it looks much better than SRV13:

https://drive.google.com/file/d/0B8_LYKyZDiajeGF2cnVibVJMMFU/view?usp=sharing

confusedgambler said:

DrDre said:

For one reason only, because that's what the colors looked like in 1977. 

Come on, Darth Vader is solid black with no detail left. You can't seriously say that looks right or good or even "what it looked like in '77" Not a chance, no way.

Also,that's not what Mike Verta's version looks like and he has better sources to judge by.

Sorry to dissappoint you, but the reference is a scan of an unfaded Technicolor IB print, corrected to the print. Although there can be variation between prints, the colors are the way they are on the print. The reason detail is lost, is because the GOUT lacks the color depth to show the detail, but on the print it is definitely there. 

For one reason only, because that's what the colors looked like in 1977. 

Here are three frames of the GOUT Star Wars, matched to a Technicolor IB print. The references came from a source that wishes to remain anonymous, and I cannot show the references themselves, but they are an exact match. The differences are interesting to say the least:

http://screenshotcomparison.com/comparison/142085

http://screenshotcomparison.com/comparison/142087

http://screenshotcomparison.com/comparison/142088

Here are some new color corrections for the video sample of the first reel of team negative1's LPP preservation. All these corrections were done with the new color correction algorithm. The model was calibrated on a single frame (8228). The model was then used to predict the corrections for the other frames. Here are the results:

http://screenshotcomparison.com/comparison/142018

http://screenshotcomparison.com/comparison/142019

http://screenshotcomparison.com/comparison/142020

http://screenshotcomparison.com/comparison/142021

http://screenshotcomparison.com/comparison/142022

http://screenshotcomparison.com/comparison/142023

http://screenshotcomparison.com/comparison/142024

Thanks! :)

In the mean time, I managed to improve the color matching method. There appears to be some color noise hidden in the raw scan, that was difficult for the algorithm to handle. The new algorithm is less sensitive to the color noise. Once again the reference is the Semi-Specialised Edition V2.2 frame posted earlier:

The results are as follows (raw/rough correction/correction with algo):

At least the waiting is over. Hope the surgery works out, such that you can finally enjoy some healthy times!