logo Sign In

Star Wars GOUT in HD using super resolution algorithm (* unfinished project *) — Page 50

Author
Time

Screenshot 3 looks almost over sharp - the other two look great :)

This is just out of curiousity - is there any way to run a halo analysis pass that finds "halos", then direct avisynth to replace only those spots with a version of your SR script that is less halo strong?

Preferred Saga:
1/2: Hal9000
3: L8wrtr
4/5: Adywan
6-9: Hal9000

Author
Time

nightstalkerpoet said:

Screenshot 3 looks almost over sharp - the other two look great :)

This is just out of curiousity - is there any way to run a halo analysis pass that finds "halos", then direct avisynth to replace only those spots with a version of your SR script that is less halo strong?

 There was some discussion a while ago about the halo's, and g-force recently suggested a solution a few posts back that seemed feasible, but also a ton of work, so if anyone wants to try they're more than welcome, but I will take a pass ;-). 

Author
Time

DrDre said:

Zyrother said:

Looks awesome! But, can we get a couple screenshot comparisons of V11 and V15? Would help to determine how much of an improvement it is.

Thanks!

 Here's a comparison between SRV11 and SRV15 (I've reduced the sharpening somewhat):

http://screenshotcomparison.com/comparison/146560

Thanks! Looks like there is a bit less noise in the darkest areas. The reduced sharpening helps too.

Awesome work! V15 looks great, and I think most people would be happy with it being the release version.

Author
Time
 (Edited)

Looks pretty great in that one screenshot, although it looks like the color or saturation is slightly less (not saying this is a bad thing).  The reduced noise seems like a good thing and it seems that the details are still there.  It also seems like with the lower sharpening that the halos aren't as obvious.  Can you post a bunch of the other standard screenshots comparing to V11?  My only concern is whether there is less 3D pop or not.  Great job!

Author
Time

thorr said:

Looks pretty great in that one screenshot, although it looks like the color or saturation is slightly less (not saying this is a bad thing).  The reduced noise seems like a good thing and it seems that the details are still there.  It also seems like with the lower sharpening that the halos aren't as obvious.  Can you post a bunch of the other standard screenshots comparing to V11?  My only concern is whether there is less 3D pop or not.  Great job!

 Here are two more comparisons:

http://screenshotcomparison.com/comparison/146614

http://screenshotcomparison.com/comparison/146616

Author
Time

It does look better on the screenshots. Less artifacts, but the details are retained.

Author
Time

Does anyone else feel like that V15 has a green tint?

There is less noise, but also less detail and the aliasing is handled worse on C-3POs shoulder... I don't really know which version I like more.

Darth Id on ‘Why “Ben”?’:

And while we’re at it, we need to figure out why they kept calling Mark Hamill’s character “Luke Skywalker,” since it’s my subjective opinion that his name is actually Schnarzle Shnuzzle.  It just doesn’t make sense!

Damn you George Lucas for never explaining why they all keep calling Schnarzle “Luke”!

Damn You!!!

Author
Time

The screenshot comparison does show some kind of color shift.

Author
Time

Looking at these on my better screen, I think SRV11 looks ever so slightly clearer and has the illusion of more detail due to the increased sharpening but it is also slightly increased halos and also the extra grain.  It would be nice to see the C3PO parts in motion with a back to back 3 second clips comparing V11 and V15.  Seeing the grain in motion will help determine whether it is best to keep it or get rid of it.

Author
Time
 (Edited)

Although super resolution is very useful, when it there is no high resolution release available, it's obviously pretty useless in the presence of a true high resolution release. Considering the picture quality of Team Negative1's latest video samples, it is pretty obvious that any GOUT upscale is going to pale by comparison. Aside from the obviously higher resolution, there's no aliasing, no halos, no static grain, and no DVNR. That only leaves the GOUT color grading, which can be reproduced as was evident from the few examples I posted of the bluray regraded to match the GOUT:

So, my question is this. If Team Negative1's preservation is as great as it appears to be from the samples, why not create a high resolution version of the GOUT from those, by color matching the preservation to the GOUT, rather than settle for an inferior upscale of those 2006 bonus DVD's? It may seem like an odd question from someone who has spent many hours working on this, but I think it is a valid question.

Author
Time

Completely valid question. And in my honest opinion, color matching Team -1s restoration is perfectly valid, and acceptable, as that has the content everyone loves. Color matching the Blu Ray with GOUT colors is also another great option too.

But, the goal of this was never to provide another preservation of the OOT, simply to prove that Super Resolution was a viable up scaling method. Which it most certainly is. Comparisons of V15 and the source are incredible.

I would completely disagree in saying that this whole endeavor was worthless. Lots of people came together, and lots of ideas were thrown around. it even spun off a deep discussion of color matching.

Whatever version of SR you choose, it is still an accomplishment that merits recognition!

Author
Time
 (Edited)

I think it is a good question. One that I probably wouldn't have brought up, due to the hard work you've put in. But if you're at the point of deciding whether or not to begin another massive encode, it's certainly worth discussing.

I admit that my enthusiasm in the end product (but not the technique) has waned a bit, now that TN1's work is shaping into something quite nice. The benefits of your upscale, to me, were:

1. It's 'pure' SW. There's no mixing of sources that need to be matched or rotoscoping to be spotted by the trained eye. This is also the main benefit of TN1. 

2. Color. It's not perfect, but it's roughly the color of the film I grew up watching. It's fairly consistent and not insane like 2004 and onwards. As you say, 35mm projects can be made to match the GOUT for those who are interested.

3. You've come up with a great script that has already been applied to other things, like Hal's project. It can be used for DeSp type projects where nothing better than GOUT is yet available.

The cons:

This is going to take you a very, very long time to start over with. You have other great projects going on. And there are two other films which will need adjusted scripts and again enormous periods of encoding.

DVNR. It's not going anywhere. 

This (I mean encoding the entire film, *not* the R&D) is being done for practical use, to view a less ugly version of the (approximately) untouched film. It's one of many options. That's what's great about this community. It has kind of a shot-gun approach. Every imaginable way of making this film exist is being attempted. That said, what is the userbase that will prefer this version to TN1 or Harmy or Team Blu? If it were a 2-day encode, I would say who cares? Let's add it to the pile and the list of options! It's not trivial when we're talking about probably more than half a year spent encoding the trilogy, if it's something five people watch once. 

I'm sure people will disagree. But I say, just get the script out there. People can do the scenes they need. Or the entire film if it's really ready-to-go. Someone might figure out how to get distributed computing power working for this, some day, and we can get it done rapidly.

I think this is worth discussing!

Author
Time

Zyrother said:

Completely valid question. And in my honest opinion, color matching Team -1s restoration is perfectly valid, and acceptable, as that has the content everyone loves. Color matching the Blu Ray with GOUT colors is also another great option too.

But, the goal of this was never to provide another preservation of the OOT, simply to prove that Super Resolution was a viable up scaling method. Which it most certainly is. Comparisons of V15 and the source are incredible.

I would completely disagree in saying that this whole endeavor was worthless. Lots of people came together, and lots of ideas were thrown around. it even spun off a deep discussion of color matching.

Whatever version of SR you choose, it is still an accomplishment that merits recognition!

 I certainly agree the journey has been worthwhile, and also entertaining. I've used super resolution to create some pretty neat looking upscales of the prequel deleted scenes for someone else's project. 

However, I must also admit, that ever since I saw -1's latest samples, and realized what's possible in the color matching department, using the GOUT as a source seems more and more illogical to me. 

Author
Time

I agree with towne32, even if you don't want to release the films themselves, releasing the script would be a major step forward in allowing other people to do their own thing, and maybe get even better results. Not only with Star Wars, but other films that have not gotten the proper treatment they deserve.

Author
Time

Zyrother said:

I agree with towne32, even if you don't want to release the films themselves, releasing the script would be a major step forward in allowing other people to do their own thing, and maybe get even better results. Not only with Star Wars, but other films that have not gotten the proper treatment they deserve.

 Furthermore, if the script is done, I don't think the person who has spent so much time on it needs to be the workhorse in producing the output. In fact, I think he's the last person who should have to take that burden.

It made sense that he was doing it before, as the script was being updated continually. 

Author
Time
 (Edited)

towne32 said:

I think it is a good question. One that I probably wouldn't have brought up, due to the hard work you've put in. But if you're at the point of deciding whether or not to begin another massive encode, it's certainly worth discussing.

I admit that my enthusiasm in the end product (but not the technique) has waned a bit, now that TN1's work is shaping into something quite nice. The benefits of your upscale, to me, were:

1. It's 'pure' SW. There's no mixing of sources that need to be matched or rotoscoping to be spotted by the trained eye. This is also the main benefit of TN1. 

2. Color. It's not perfect, but it's roughly the color of the film I grew up watching. It's fairly consistent and not insane like 2004 and onwards. As you say, 35mm projects can be made to match the GOUT for those who are interested.

3. You've come up with a great script that has already been applied to other things, like Hal's project. It can be used for DeSp type projects where nothing better than GOUT is yet available.

The cons:

This is going to take you a very, very long time to start over with. You have other great projects going on. And there are two other films which will need adjusted scripts and again enormous periods of encoding.

DVNR. It's not going anywhere. 

This (I mean encoding the entire film, *not* the R&D) is being done for practical use, to view a less ugly version of the (approximately) untouched film. It's one of many options. That's what's great about this community. It has kind of a shot-gun approach. Every imaginable way of making this film exist is being attempted. That said, what is the userbase that will prefer this version to TN1 or Harmy or Team Blu? If it were a 2-day encode, I would say who cares? Let's add it to the pile and the list of options! It's not trivial when we're talking about probably more than half a year spent encoding the trilogy, if it's something five people watch once. 

I'm sure people will disagree. But I say, just get the script out there. People can do the scenes they need. Or the entire film if it's really ready-to-go. Someone might figure out how to get distributed computing power working for this, some day, and we can get it done rapidly.

I think this is worth discussing!

 I think you hit the nail on the head. I agree completely. I've cleaned up the SRV11 script for SRV15, so I will definitely post that later today, when I get home, but the encoding takes an terribly long time, and most of it's to do with reducing all the artifacts that are caused by applying super resolution to a terrible source like the GOUT, not so much the super resolution itself. 

Author
Time

Awesome.   I'd love to take a crack at encoding this myself.  Just out of curiosity,  how specific is the script to Episode IV?  Can it be easily modified to work for Empire and Jedi?  

If I had some gum, I’d chew a hole into the sun…

Author
Time

g-force said:

My opinion:

1. Tech IB print has colors that differ from the prints on opening day. They are all there, with little fading, but that doesn't mean they are the same as what was show at most theaters.

2. The scan that Harmy worked from had colors that didn't quite match the Tech IB.

3. Harmy merely approximated the colors of the scans he had. Sometimes to varying degrees of success.

So, that's 3 levels of wrong.

I think you're better off looking at something like the JSC, but again, that was fucked with as well.

Anyone looking to get the "correct" colors is on a fools errand. They no longer exist. There is no reference that I know of that is representative of them. Their fire has gone out of the universe. :)

-G

 I would agree with you that none of those is a gold standard to try and match. That said, I think there is a way to come as close as possible with what we have available. The question is what to use as a guide and on that people will differ. Some will point to the how little technicolor fades. Some will point to the older editions made from original IP prints. I prefer to rely on the other movies, behind the scenes photos and film, to uncover an image of what the original colors were and try to mold the film to match.

That said, this thread is really about heightening the resolution of the 2006 DVD versions of the films (apparently taken from the DE LD master tapes). But that edition really had the best colors of all of them and is a good landmark to go by.

Author
Time

To me, it's all about the journey and not so much about the end product, although it's nice to reach the end.  Can't multiple people do the encode in 10 minute chunks? Then agree on the best way to encode it so that one person can take all of the files and do one final splice together.

On a side note, I know we are all waiting and hoping for a perfect restoration of the original trilogy.  If we ever actually got it, the whole point of this site would sort of go away.  No more journey.  Although again, it would be nice to finally get it someday.

Author
Time

thorr said:

On a side note, I know we are all waiting and hoping for a perfect restoration of the original trilogy.  If we ever actually got it, the whole point of this site would sort of go away.  No more journey.  Although again, it would be nice to finally get it someday.

 Hah! You're suggesting that the members of this site would ever be satisfied with the film. :P

Author
Time
 (Edited)

Even if we got a 'perfect' restoration, there would still be people here who would value a GOUT upscale simply because of nostalgia. Besides, there ARE differences between the '77 version of the film and this one, which would be valuable on their own.

About v15: It looks good, I like it more than any previous version! The problem has always been the ringing artifacts, and these seem to be impervious to any attempt to bring under control. But perhaps if there was an automated tool, such as Photoshop's Magnetic Lasso tool, which could identify edges of a sufficient contrast (such as Vader's helmet against a featureless wall) it could be made to equalize lesser variations in luminosity which run parallel to these edges, but only equalize them on the side of the edge which has the most parallel gradients (which are the ringing artifacts). In this way, details on the other side of the edge remain unaffected, which is where most of the superresolution detail is gained.

You probably don’t recognize me because of the red arm.
Episode 9 Rewrite, The Starlight Project (Released!) and ANH Technicolor Project (Released!)

Author
Time

The end result of GOUT in this project is not worthless or anything close to that; in fact, it'll be the best "original" print perhaps for several years from now on. TN1's version may take years to be released, and Disney may never even release their's. The logical thing would have been to do it prior SW VII. It didn't happen.

Even when this GOUT upscaling becomes obsolete one day, it will be still useful as a guide for other projects. Probably it'll even have intact frames, details that are not present in other prints due to damage.

thorr said:

If we ever actually got it, the whole point of this site would sort of go away.  No more journey.  Although again, it would be nice to finally get it someday.

 

Oh, don't worry, there'll be plenty of other journeys to make... with different but still important movies.

Author
Time

Here's the script for SRV15:

orig=AviSource("SW.avi")

orig=Trim(orig,0,50000)

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)

sr=Average(sr1,0.25,sr2,0.25,sr1r,0.25,sr2r,0.25)

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)

edi=Median(edi,q1,q2)

edi=ConvertToRGB24(edi)

s64=ConvertToRGB24(s64)

sr=ConvertToRGB24(sr)

sr=Average(sr,2,s64,-2,edi,1)

edicd=ConvertToYUY2(edi0)

edicd=ConvertToYV12(edicd)

edicd=ConvertToRGB24(edicd)

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

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

Spline64Resize(1920,816)