Why does the restoration of 'The Third Man' look weird? | Hacker News Reader

Why does the restoration of 'The Third Man' look weird?

blouinartinfo.com

35 pointsanigbrowl46 comments

"Why does my favorite old movie look like shit? I want 1100 words by Thursday"

"Because they're compressing it too much. There are parameters in all compression formats where you can make the files bigger or smaller, and they're making the movie too small, so a videophile will notice differences. Resolution is the largest of these parameters, and it corresponds to actual hardware, but it's not the only one. They're cheating you out of accurate movement in order to make their data processing easier."

"I do not understand the words that are coming out of your mouth because I have an art degree. 1100 words. Emotive language. Thursday."

-----

You have a problem with a transfer? Reduce the amount of compression to well beyond the level your focus group can detect, and increase the resolution. Don't talk about “the random scatter of photosensitive molecules in the emulsion creates an overall softness and shimmer,” because data is data, and sufficient amounts of data can represent that scatter just fine.

From MPEG-2 to H.265, video format advances have reduced data needs to maybe 20% of what they used to be, while from 480p to 8K, resolution improvements have increased data needs by a factor of 108x. Businesses have failed to cope honestly: They're selling the higher resolution but compressing more heavily so it looks just like it used to look, or giving only token data allowances so 1080p is given 50% more data than 480p rather than the 575% it needs in order that a 480p inset in that larger picture has the identical performance as the lower resolution stream would.

And that is why Youtube looks like shit, and that is why film transfers are still not performing at the level we want.

I'm not sure why are you talking about H.265 at all. DCP format uses JPEG 2000 compression for video, and it doesn't use 1080p: the screen size are from 1998×1080 to 4096×2160.

h.265 is a generic example of the state of the art in video compression. The distinction between 1080x1998 vs 1080x1920 is negligible. The point stands without the minutia.

When you double the resolution, the amount of data you're representing goes up by a factor of four. Businesses who double the resolution of their streams have not raised their data rates by a factor of four or even a factor of two, and the progress in encoding, while significant, has not been remotely sufficient. We're compressing more heavily than ever before because for some reason marketting can sell resolution at the edge of theoretical human perception, but not a fat data pipe well within human perceptual range.

You're not wrong, but you have to be a bit careful with this line of reasoning when talking about compression, particularly: 4x the content in terms of pixels is not necessarily 4x the information content, and the latter is the real issue.

Also worth thinking about. With transferring old film stock to digital, you have a potentially long pipeline of which compression is just that last step. I'd hesitate to blame all visual distortion (relative to the original) on compression without looking at it carefully.

The standard for HD television in the US, ATSC A/53, provides for:

"The maximum bitrate of the MPEG-2 video stream is exactly 19.4 Mbit/s for broadcast television, and exactly 38.8 Mbit/s for the "high-data-rate" mode (e.g., cable television). (The practical limit is somewhat lower, since the MPEG-2 video stream must fit inside a transport stream, with overhead, sent out at 19.3927... Mbit/s for broadcast.)"

38.8 megabits per second is 291 megabytes per minute, or 17.5 gigabytes per hour. You're more likely to be watching something that's compressed an order of magnitude more, because businesses think they can cheap out on it without consequences. Here's Youtube's bitrates: http://www.sciencemedianetwork.org/Blog/20130624_YouTube_bit... .

Transferring analog content lossfully at similar resolution is another issue - if the VHS tape has about 240p equivalent resolution, and you transfer it with perfect fidelity to an uncompressed 240p digital stream, you're losing a lot of information, because those dots don't line up perfectly with those pixels, instead smearing out over multiple pixels. Instead of thinking 'equivalent resolution', think Nyquist frequency.

If you've ever seen OTA HD at full bandwidth, it's beautiful.

Sadly, nobody ever sees it because either a) the broadcaster has added 7 subchannels and/or b) almost everyone is watching on cable or satellite and the provider has compressed it all to hell to fit their own transport.

I think Fox is the rare broadcaster in my city that doesn't subchannel and puts sporting events on in full bandwidth mode.

It's not entirely about compression - some "remasters" have been de-noised (i.e. no more grain - which in my opinion (and it is opinion) loses some of the atmosphere of a film (hell, if grain is unimportant, why do so many games have shaders which add it?)), some are run through a sharpen filter to deal with out of focus scenes and to "improve" the image quality, and so-forth - so it's not so much about compression as it is about other subjective "improvements" made to the transfer.

An example of something which has been done wonderfully, again in my opinion, is last year's restoration of "The Cabinet of Dr. Caligari".

Interestingly, they re-added the original tints to the remaster, as what's not commonly known is that many "black and white" films were projected through a coloured glass slide, in order to either give the entire scene a hue, or even to make the ground red/brown and the sky blue.

"Interestingly, they re-added the original tints to the remaster, as what's not commonly known is that many "black and white" films were projected through a coloured glass slide, in order to either give the entire scene a hue, or even to make the ground red/brown and the sky blue."

I'd always wondered where Top Gear gained its impressive cinematic style - they must have happened upon a treasure trove of those two-color filters, and asked what it would feel like applied digitally to every other scene with the saturation pumped up.

And then they went crazy with it, to the point where you were watching a purported car review that successfully concealed the color of the vehicle by the application of art. It became distracting.

But not as distracting as Top Gear's urge to rotate the camera to emphasize the slope when a car is going up or down hill. Or HGTV's urge to run segments containing fireplaces in reverse.

You can still buy such things - I use them in photography - See Cokin, for instance - it's their entire business. http://www.cokin.co.uk/

> (hell, if grain is unimportant, why do so many games have shaders which add it?)

I seem to recall there even being an added grain parameter in the h264 spec!

> (i.e. no more grain - which in my opinion (and it is opinion) loses some of the atmosphere of a film

I hate film grain. I don't like flat images either, but I'll take them over film grain. When I look at grass I don't see dots randomly moving all over the place, so I don't want to see them in a film either.

Maybe you prefer the look of video. I very much prefer the look of film (tough I use both) but not because of fidelity. It's like a waxwork in Madame Tussaud's is more lifelike than a stone or bronze statue, but I prefer the latter.

Is "I do not understand [you] because I have an art degree" the same as "you would benefit from learning how to communicate effectively with a non-technical audience"? Probably not, but they seem to be conflated.

That's not to defend poor journalism. But it is important to notice where willful ignorance or aggressive blaming damages communications.

Compressed data bandwidth for a video does not scale with resolution. i.e. quadrupling the number of pixels does not require quadruple the bandwidth for the same quality.

I believe it goes up by the square root. So quadrupling the number of pixels needs only double the bandwidth.

There are several reasons for this, one of which is that motion compensation is constant - it doesn't much matter what the resolution is, it takes the same amount of data to encode motion. Another is that as resolution goes up there are more opportunities for similar areas in the image which compress well.

If you are ever in Vienna I highly recommend taking the underground Third Man sewer-tour and then afterwards go see the film at the Burg Kino which screens it daily.

https://www.drittemanntour.at/en/index.html

http://www.burgkino.at/film_detail.phtml?fil_id=4&per=now

As a Viennese who took that tour together with some international guests, I highly recommend _not_ doing it, because you're not allowed to see very much. I found it quite disappointing.

You recommend doing nothing over doing something quite interesting?

I've done it three times with people visiting me. Not everyone wants to spend half a day underground.

The point about grain resonated (heh) with the article from last week about Stochastic Resonance ( https://medium.com/the-craftsman/the-curious-phenomenon-of-s... )

Anyone who retouches photos knows that a touch of grain can hide a multitude of sins. I used to keep a high-res scan of film grain around for just this purpose.

This isn't a problem with technology - it's a problem with taste. The people overseeing the digital transfer either don't share the taste of the author, myself and many others, or arguably, simply lack taste.

I'm pretty sure that article is wrong, according to the article Stochastic Resonance is about amplifying non-linear signals by adding noise. The added noise amplifies the signal (i.e. it the result is more than the sum of the weak signal and the added noise).

What he has done is simple dithering (https://en.wikipedia.org/wiki/Dither).

> Dither is an intentionally applied form of noise used to randomize quantization error, preventing large-scale patterns such as color banding in images.

Dithering works on standard linear processes (like adding noise to images).

Here's the thing, motion smoothing fakery is for sports. 24 fps is what we have mentally trained ourselves for our movie-going experience. Change that and even the casual viewer who has no idea what motion interpolation is will feel "something just doesn't feel right".

In my opinion, there are three major problems: overzealous sharpening, different light ratios and the frame rate dependent flickering.

I consider first two absolutely unacceptable and I've seen enough "restorations" for which I wondered if anybody actually looked at the results or just passed the thing through the machines. But I prefer the stable picture even when the result appears less "magical."

The worst of it is that the sharpening algorithms used are total garbage. You’d think with the the money being thrown at this stuff, they could hire a couple of image processing experts to advise them.

Given the subject matter, the aspect ratio of the shot at the top is surprisingly inauthentic!

YouTube trailer - which gives away a couple of good scenes - viewable at 1080p: https://www.youtube.com/watch?v=F-QWLAndD1E (I actually think it looks quite good! - though maybe it won't work on a cinema screen)

Is it just me, or is there an enforced newsletter sign-up modal when you land on that page?

I can see the same thing and there is no button to close it.

https://i.imgur.com/PjBQCfm.png

There is a button in the top-right, you just can't see it.

sadly, not just for you

There was a black 'x' in the top right of the pop over for me. Are you using a script blocker or something that's preventing the button from loading?

Oh, thanks - that is why I didn't see the 'x'. Well, I just deleted the modal dialog from the DOM... No I'll have to add "mc_embed_signup" div to my adblock blacklist.

In Firefox, right-click on the modal dialog, select 'Inspect element', find <div id="mc_embed_signup" style="">, right-click on it to select 'delete node'... Modal dialog is gone - you are in control of the rendering !

2k cinema is the same as a dad? That's bullshit. I watched Sukiyaki Western Django at a film festival and there had been a cock-up with the movie, so it was still in a different country when it was supposed to show. They ended up using a dvd as a source for the projection in the cinema.

It was god damn awful.

I think he means that 2k cinema is the same as dvd in a tv/computer screen not in a movie theater screen

And he's still wrong even if he meant that.

A DVD is approx 0.7K - maybe a bluray could be considered 2K, but not a DVD.

It's because of the size and distance comparisons, its not about the resolution. His point is that a 2k cinema movie is like watching a DVD vs say a bluray or an 8k cinema movie.

That, unfortunately, is what's horrible about the article -- the author never makes clear whether this is an objective or subjective comparison. Is the image too poor, or too "good"?

Does anyone know who did the restoration work on The Boy in the Plastic Bubble? (The made for TV version, with Travolta, and Mr. Brady--197?)

I just saw a restored copy on one of the cable channels, and even though it was not a full screen copy; it was watchable.

I liked the original movie. It was shot of video tape, and was made for T.V. movie. I don't even think they bothered to copyright it? The original copy(master) degraded over time, and for years I thought it was lost.

I'm can't find any info on the Internet on who did the work? It's not a theater quality restoration, but I'm suprised anyone could restore that old, degraded video.

In America, and every other signatory to the Berne Convention, creators automatically own the copyright on whatever they create. They would have to explicitly waive all of their rights before it would be un-copyrighted. https://en.wikipedia.org/wiki/Berne_Convention#Content

"Skin tones look polished and movement is stabilized, giving the image a sharpness that was not originally there."

That actually sounds wonderful to me! The Third Man is probably my favorite film. I always noticed the jitter during the opening credits (zither strings twitching around the frame) and I'm pretty sure Carol Reed would have preferred it to be steady there. I certainly will. I can't wait to see this. As for film historians, hopefully the digitized source will be available somewhere, so they can have the scratches and jitter, if they want them.

For anyone with an interest in the rise of digital over film in the movies, watch "Side by Side" (http://sidebysidethemovie.com/). It's on Netflix streaming. It's an excellent documentary that interviews many different directors (some who prefer digital, some who prefer film), explains the technology behind both, gives a history of the rise of digital and in general just gives you a greater appreciation for movies.

And it just might make you respect Keanu Reeves.

Can anyone else see how to dismiss the lightbox?

(How do we get Google to penalise lightboxes?)

On firefox, the close button at top right doesn’t seem to show up if you have the ad-tracking blocker turned on but if you mouse over the top right corner the pointer will turn into a hand denoting the presence of a link & you can click there to close the dialog box.

right click, "Inspect Element". Add an extra CSS rule to the element{}, of display:none. This works on almost all sites with popup overlay crap.

There's probably a plugin which lets you do this with a single toolbar button...

For sites that do this, the appropriate single keypress is Ctrl-W.

Some sites appear to block Ctrl-W these days. Those sites get the full AdBlock "Go Away" treatment.

If they're worried about people throwing the word "film" around too flippantly, they probably shouldn't do the same with "celluloid". Acetate, please.

Reply on news.ycombinator.com