TPM 1080p Theatrical Preservation (a WIP) — Page 4

Harmy said:

I don't think pan and scan VHS has better resolution than widescreen LD, let alone digital PAL anamorphic broadcast.

Oh the vertical resolution would be better, but the horizontal fidelity would probably be no better. The thing with all analogue video is that they all have distinct lines of resolution - NTSC is 480 lines* regardless of whether it's VHS, Betamax, or LD. Note there is actually 525 lines of resolution in the NTSC standard, but the picture window is 480 lines.

The horizontal information is compressed, when decompressed it yields 720 pixels (or whatever amount your TV can handle). The actual amount stored on VHS is said to be only around 320 distinct points (pixels) or so. Also expressed as 240 lines per picture height (240 * 1.33333 = 320). Laserdisc, I believe, yields perhaps 80-90% more horizontal fidelity. And DVD yields more still (no horizontal compression).

And that's just luma - chroma is compressed even more.

And then there's the noise. VHS is noisier than LD which is noisier than DVD (no analogue noise).

I've never seen the officially released HK TPM VCD, I'd be curious to see it though (so if anyone has it - spleen it!!) VCD has only 50% of the vertical resolution, and its horizontal resolution is comparable to VHS (Luma that is, Chroma is better).

Upscaling...

As I mentioned earlier in the thread, it wasn't easy designing an upscale script for this film. Not only that, but the upscale itself is now extremely slow - making progressing it further difficult. Here's a comparison to my old upscale script from last year (there are two scenes in that link).

Here's the links for LD and TB:

Laserdisc vs Upscale (4 scenes in the link)

TB vs upscale (4 scenes in the link)

Your upscale script works very well - also thanks to the fact you use two sources; the comparisons with each sources speak for themselves!

The new script is not that different from the old one (which was good!) but if you look close, yes, the difference is there, not that great, but the new one improves a bit the very good result of the old script!

Thanks. I wish it looked better though. The crowd shot is one of the better looking ones. Andrea has kindly offered to provide better quality caps of the LD scenes, so once I have those the scenes may look a bit better than they presently do.

Since I haven't posted one in a while, here's a preview clip. It shows the LD+DVB upscaled crowd shot and the DVD-extras sourced "Sebulba" scene with Watto and Warwick Davis cheering Sebulba.

Also, if you're curious you can download the Sebulba scenes and compare them.

Downloads:

Preview clip (52.3MB).

Sebulba scene - DVD version:
1. Uncompressed AVI (96.8MB)
2. x264 crf16 (2.56MB)

Sebulba scene - LD+DVB version:
1. Uncompressed AVI (96.8MB)
2. x264 crf16 (2.34MB)

You'll notice that the version in the preview clip has been softened slightly and has a grain-plate added, compared to the raw version.

Maybe it's possible to get the best quality of those scenes using my captures, RU.08 align script and DrDre SuperResolution upscale... we should get the best result ever!

What do you think?

As DrDre already discovered, the Infognition plugin probably isn't suitable for this source:

Geometrical SR reconstruction algorithms are possible if and only if the input low resolution images have been under-sampled and therefore contain aliasing. Because of this aliasing, the high-frequency content of the desired reconstruction image is embedded in the low-frequency content of each of the observed images. Given a sufficient number of observation images, and if the set of observations vary in their phase (i.e. if the images of the scene are shifted by a sub-pixel amount), then the phase information can be used to separate the aliased high-frequency content from the true low-frequency content, and the full-resolution image can be accurately reconstructed.[12]

In practice, this frequency-based approach is not used for reconstruction, but even in the case of spatial approaches (e.g. shift-add fusion[13]), the presence of aliasing is still a necessary condition for SR reconstruction.

(Wikipedia)

The reason why it works for the GOUT is because of how it was transferred to video. The aliasing is necessary as there is meaningful information missing between the scan-lines that can be recovered from from other frames. It'd be nice if Infognition was more forthcoming with this information.

In any case, to use SR you would need to do the upscale first and then the alignment.

Well, if the GOUT benefits of SR, laserdisc will too; as I tested myself with the '97, SR squeeze some details more than non-SR upscale filters; also, DVB probably will benefit, too, as supposedly its master was HD and downscaled to SD...

The fact the upscale is made first is not a problem; the hard work is to find the perfect settings for the alignment!

Have a look if you like, here are all the scenes with just a crop and quick transcode (CRF-10). I've also put uncompressed AVIs up.

LINKS MOVED (see the op).

I can't get SR to find meaningful detail in there, but perhaps DrDre can.

I've just cleaned-up the op now as well. All these links and more are in the op.

If compared side by side, DVB+LD is clearly less detailed than DVD, in particular when seen frame by frame; but in motion, difference is not that great, and I think these shots could be inserted in the movie without keeping out too much the attention of the viewer due to lower resolution.

Thinking about the quality of each source used, the final result is really great, not so distant from the DVD... good job!

I'll capture the needed shots ASAP - hope today... I'll keep you informed in this thread.

That should would work fine as it is, yes. Just as the crowd shot does - it doesn't look out of place at all (have a look at this comparison to the true HD version!). The senate is one of the lower quality ones. Although, it doesn't look that bad in motion either.

But if I could comment on your earlier reply where you said "the best result ever" - perhaps from the sources I have, yes. But sometime in the future someone will scan those scenes in true HD from a print, and that will be the best result.

WOW! Very similar indeed!!!

And, when I wrote "the best result ever", I forgot to add "until now"...

Where's that confounded bridge?
Left is Handman's VHS, center althor1138's LD, right adywan (anamorph.)

It's part of my latest frame-counting exercise
sample

VHS is most complete, missing only 1.5 frames,
althor missing 6 frames,
adywan missing 5 frames, with one extra frame that doesn't belong.

Yes but the Z-VCD clearly cuts out earlier than the LD - and you can tell this is the case by the position of the arch in the foreground. The LD has about 11 more frames than the Z-VCD.

There's only one LD release of TPM though.

Okay so Andrea has very kindly sent me some caps. I've done a quick manual IVTC and assembled them together. As I don't want to deprive anyone of the opportunity to have a go themselves at upscaling them, I'm posting them here - note that the alignment script needs to be modified to work with Andrea's. I don't have the time right now to do that, or to work on the upscale script. Therefore there are now 3 sources for these scenes.

These are all the scenes requiring upscales. The x264 are transcoded with a crf setting of 10, or you can download the (uncompressed) AVI files.

• DVB recording (x264) / Lossless AVI (TB source)
• LD rip (x264) / Lossless AVI (Althor)
• LD rip (x264) / Lossless AVI (Andrea)

The first part of the upscale is the alignment script, since these videos were already cropped to use it you just need to remove the first crop from each source.

Oh, then I'm most interested in seeing it. Can you put it on Myspleen?

Here's a quick compare of a shot that was changed slightly for the bluray. More endless tinkering from Lucas.

http://screenshotcomparison.com/comparison/133266

RU.08 said:

Here's a quick compare of a shot that was changed slightly for the bluray. More endless tinkering from Lucas.

http://screenshotcomparison.com/comparison/133266

Resized hdtv pic for easier comparison - http://screenshotcomparison.com/comparison/133303

Yes, there seems to be added reflections in the windows for the blu. What is that shadow near the bottom middle window post in the hdtv? Doesn't seem to be on the blu.

Yeah, as you mentioned there are a couple of differences in such an ordinary shot. The windowpane placement is slightly different, there's glare/reflection in the windows, and that "shadow" looks to be a part of the window-frame, and is gone for whatever reason in the bluray.

Also, this is one shot where the difference in quality between the HDTV and the bluray are quite obvious, in most other shots this is not so much the case.

I don't know, I just don't know.

Apart from the different cropping and extreme contrast difference, I see no changes.

Every element seems to be in exactly the same place, even the windows in the foreground. I believe things just appear to shift because of the extremely heightened contrasts in the HDTV shot blowing out the edges of the elements.

That one shadow difference along the bottom is the only thing I can't account for.