Best way to convert 1080i to 1080p?

[jrodefeld]

I'm working on a fan restoration of John Woo's The Killer and I'm hoping you could help me out with something.

The highest resolution version available is probably Dragon Dynasty's Blu-ray. The thing about that version is that it's encoded as 1080i and it's riddled with artifacts and jittery motion. Apparently they used a PAL source and didn't bother to properly convert it.

There's another guy who used to be a member of AsianDVDClub (before they shut down) who apparently was able to fix many of the problems with this transfer.

This is what he said about it:

"I deinterlaced and decimated the Dragon Dynasty BD to a progressive 23.976 fps. I'm surprised at how good it turned out. For a while I thought the Dragon Dynasty BD was completely unsalvageable.

A few notes though; It's not perfect. There's some ghosting and bad frames (though, it's hardly noticeable during playback), and the grain still looks unnatural."


I'll see what I can do about the ghosting and bad frames later, but I first want to get the best quality 1080p 23.976 video I can with the smoothest motion and least artifacts.

Is it as simple as a script like this?

QTGMC(Preset="Slow")
Tdecimate()


I uploaded a clip so you can see what the original looks like:

https://www.dropbox.com/s/mvqrzygnfrhppbr/TK_Snippet.ts?dl=0

I used a program called VideoReDo to cut the original M2TS file without reencoding so it should look identical to the original file.

I'm competent with Avisynth but I'm by no means an expert. So I like to get second opinions about scripts I use from people who know more than I do.



Thanks for the help!

[poisondeathray]

Field blended from 25fps to 29.97i . It should be 25fps


AssumeTFF()
QTGMC()
SRestore(frate=25)

[Cary Knoop]

If it is originally a 24p movie and the person putting it on BR knew what he was doing you should not deinterlace it, instead, you need to perform an inverse telecine.

But then there are hackers who blend or destroy the movie in other ways so in this case you have no other choice than to deinterlace and srestore().

Like so: https://www.dropbox.com/s/8nmggigp8s618xt/snippet.mp4?dl=0

[manono]

..but I first want to get the best quality 1080p 23.976 video I can...

Yes, field blended from a 25fps source. No, your own script is no good (and probably the script of the guy that told you what he did).

To return it to film speed add an:

AssumeFPS(23.976)

after pdr's script. Or even AssumeFPS(24) as 24fps is perfectly okay for Blu-Ray. You'll also have to slow the audio to match.

I think you'd be better off acquiring the Hong Kong version:

https://www.blu-ray.com/movies/The-Killer-Blu-ray/23245/

Use field blended garbage only if there's nothing better out there.

[jrodefeld]

If it is originally a 24p movie and the person putting it on BR knew what he was doing you should not deinterlace it, instead, you need to perform an inverse telecine.

But then there are hackers who blend or destroy the movie in other ways so in this case you have no other choice than to deinterlace and srestore().

Like so: https://www.dropbox.com/s/8nmggigp8s618xt/snippet.mp4?dl=0

Okay, I'm a little confused now. This is from a commercial DVD put out by the company Dragon Dynasty.

I'm pretty sure they used a 25fps PAL source, not a 24fps source.

Using the script poisondeathray recommended above gave a pretty good result to my eyes.

If I have to use QTGMC should I keep default settings or could I fine tune the parameters? I'm only interested in the best quality result, I don't care how long it takes.

[jrodefeld]

Yes, field blended from a 25fps source. No, your own script is no good (and probably the script of the guy that told you what he did).

To return it to film speed add an:

AssumeFPS(23.976)

after pdr's script. Or even AssumeFPS(24) as 24fps is perfectly okay for Blu-Ray. You'll also have to slow the audio to match.

I think you'd be better off acquiring the Hong Kong version:

https://www.blu-ray.com/movies/The-Killer-Blu-ray/23245/

Or even using the Criterion Collection DVD as a source. Use field blended garbage only if there's nothing better out there.

The Hong Kong version is probably worse, if you can believe that.

It's the correct framerate and doesn't have all the motion problems this one does, but it's still a very poor upscale with substantially worse detail.

Sad to say this one might be the best source I've got at the moment.

Is there no way to adequately fix a field-blended source?

[manono]

Is there no way to adequately fix a field-blended source?

poisondeathray already gave you the script. Why are you asking again?

I edited my previous post to remove mention of the Criterion DVD after learning it sucks.

[jrodefeld]

poisondeathray already gave you the script. Why are you asking again?

I edited my previous post to remove mention of the Criterion DVD after learning it sucks.

I know that. It's just that your post gave the impression that even the "best" way wouldn't give a very good result.

But, it's the best I've got so I'll do what I can with it.

Yeah, the Criterion DVD sucks. It came out in like 1997, is non-anamorphic and there's tons of dirt and dust all over the print they used. Detail is very poor. Colors are good, but that's the only thing you can say in it's favor.

[manono]

It's just that your post gave the impression that even the "best" way wouldn't give a very good result.
I see. You'll get the best possible result when starting with a field blended source. But they had a nice hi-def master, even if it was 25fps. There were other ways to make the blu-ray without resorting to field blending. It was done by incompetent people that just didn't care.

The Hong Kong one was upscaled from standard-def? What about the Japanese blu-ray? Do you know?

[Boulder]

If I have to use QTGMC should I keep default settings or could I fine tune the parameters? I'm only interested in the best quality result, I don't care how long it takes.
You could try this:
QTGMC("very slow", search=5, searchparam=4, pelsearch=4, dct=5, tr2=0, chromamotion=true, chromanoise=false, SourceMatch=2, Lossless=2, EZKeepGrain=0.4, Sharpness=0.1)

It will keep the denoising part of QTGMC quite minimal, I like it better than the default values.

[Mounir]

the "very slow" preset is not minimal.. Minimal would be "fast"

[Gser]

the "very slow" preset is not minimal.. Minimal would be "fast"

He meant that the amount of denoising caused by the deinterlacing process will be minimal.