The Hobbit - 48fps

[2Bdecided]

CRTs use phosphorescence afterglow (aka. ghosting) to show one field while displaying (writing the pixel) the next field, which is exactly a kind of blend.They really don't. It's more than 90% decayed by the time the next field is drawn.

...and I can't believe you're arguing with Didée about interlacing...!

Cheers,
David.

[scharfis_brain]

watching interlaced crt clearly Shows stairstepping But almost no motionblur. So no field Blending on native interlaced crt occurs at all.
And for sure on modern LCD or LED TV with a decent deinterlacer built in will show all much better image quality than the native interlaced crt.

[Shinta]

But the figure of 70% I gave has nothing to do with actual numbers of pixels. It's a quantification of the perceived loss of vertical resolution when comparing interlaced and progressive at a fixed resolution with the human eye.

Since when is someone able to measure 'perceived' something? And isn't 'quantification' just another word for 'educated guess'? ;-)

Of course one can always factor in some human psychology and physiology, not to mention the (bad) resolution of the human eye...

And always remember: Please watch this show in a brightly lit room and don't sit to close to the screen.

[wonkey_monkey]

Because, why on earth would you record in 48p and then convert to 48i (sped up to 50i)? Doesn't sound that reasonable IMHO.

The source is progressive, so why not keeping it like that?

All sources, ultimately, are progressive. At the risk of repeating myself, interlacing, at the cost of a few minor and occasional artefacts, saves you 50% on data at a loss of just 30% (and with modern deinterlacers, much less) in terms of quality. The video compression world would be very interested in any new method that could do the same again. It also helps that deinterlacers are already everywhere, ready to be taken advantage of.

Since when is someone able to measure 'perceived' something?

You show two sources to someone and ask them which is better. Alter the sources and repeat until most people agree that they look about the same. And that's just the non-technical method - you can measure deinterlacer quality with all kinds of metrics.

David

[nevcairiel]

All sources, ultimately, are progressive.

While the real world certainly is not interlaced, there are plenty NTSC cameras that directly record interlaced material, so unless you count the photons in the real world, the source could be considered interlaced, or something... :p

[wonkey_monkey]

so unless you count the photons in the real world

That is what I meant :D Real-world photons (or the human perception of same) are what video is (usually) aiming to represent.

[kolak]

Discussion regarding BD is bit pointless, as no professional studio will make 720 50p or some fancy pulldown 60i disc for the title like Hobbit- it simply won't happen (on top of that having already specially made 24p version).

Not saying that it's better but I like 720 50/60p sources and hope interlaced will be finally forgotten- it's time to forget it- technology is good and cheap enough.

[Lyris]

Also, some users here were arguing that 50 fps (or 50 Hz) might not be widely supported in some NTSC regions.

But, ... seriously, most HDTVs sold in the last five or even more years should support 50 Hz, don't you think ?

No. Almost all TVs in the US will refuse to show 50hz signals. I was one of the calibrators / speakers at the Value Electronics Shootout last year in New York. I'm from Europe so brought some 50hz test content with me.

Out of the 6 flat panel displays there, an LG Plasma and Sharp ELITE LCD showed the 50hz content. The Panasonic LCD, Panasonic Plasma, Samsung LCD and Samsung Plasma refused to show anything. Kind of crazy in 2012...

Even if they could, support for the 50hz frame rates is optional for BD players sold in 60hz countries. Fewer BD players in the US can play 50hz.

[Sagittaire]

Well it's possible to make objective comparison with 1080p50 source at BD50 bitrate like that:

1) 1080p50 -> splineresize 720p50 -> H264 compression at average 30 Mbps -> splineresize to 1080p50

2) 1080p50 -> interlacing with splineresize at 1080i25 -> H264 interlacing compression at 30 Mbps -> deinterlacing at 1080p50

Anyway no doubt that 750p50 will be really better than 1080i25 even with the best possible deinterlacing filter simply because H264 interlaced compression is really hard (50% advantage for progressive over interlacing for same source). For same 720p50 pixel quality (aka quantizer) at 30 Mbps, you must use 45 Mbps for 1080i25 encoding.

[wonkey_monkey]

1) 1080p50 -> splineresize 720p50 -> H264 compression at average 30 Mbps -> splineresize to 1080p50

TVs have much better resizers than that!

simply because H264 interlaced compression is really hard (50% advantage for progressive over interlacing for same source).

What do you mean about interlaced compression being "really hard"? And what do you mean about "50% advantage"?

[2Bdecided]

But is there any such source material available somewhere for testing?You show your ignorance. The EBU has been doing these tests for years.

You can start with 4096x2160p50 content if you wish...
http://tech.ebu.ch/testsequences
...or use the legacy(!!! :) ) 1080p50 streams...
http://tech.ebu.ch/Jahia/site/tech/cache/offonce/hdtv/hdtv_test-sequences
(note ftp access on the right)

Cheers,
David.

[2Bdecided]

Anyway no doubt that 750p50 will be really better than 1080i25 even with the best possible deinterlacing filter simply because H264 interlaced compression is really hard (50% advantage for progressive over interlacing for same source). For same 720p50 pixel quality (aka quantizer) at 30 Mbps, you must use 45 Mbps for 1080i25 encoding.

Those would be laughable bitrates in today's broadcasting world.

You do realise that they sell 1080i25 BluRays and people are happy to accept them as substantially artefact-free?

Cheers,
David.

[wonkey_monkey]

"In my experience, (pure interlaced) 50i/60i sources and their (bob-deinterlaced) 50p/60p products, result in similar bitrates/file sizes when encoded at the same Constant Quality i.e. interlaced encoding is around 50% less efficient."

That seems a pretty big conclusion to draw from sparse evidence. Who's to say exactly how Constant Quality works, for one thing? I just tried something similar, and although I did get approximately the same filesizes using interlaced or progressive, the interlaced version showed considerably less blocking. If it was 50% less efficient, I'd expect it to show the same amount of blocking.

David

[xooyoozoo]

Most or all of those sequences were shot on an Arriflex 765.

[xooyoozoo]

The 'quality' of those sequences, as defined by the transformation of light into pixels, is frankly not up for debate.

What you seem to be concerned with is the visual 'pleasantness' of the clips, which is a different matter and discussion altogether, as it deals with postprocessing and color grading. There are perhaps better avenues (http://instagram.com/) to satisfy your wishes.

[2Bdecided]

The 'quality' of those sequences, as defined by the transformation of light into pixels, is frankly not up for debate....and even more relevant to a discussion of the bitrate and quality trade-off of various scanning resolutions and modes, they were designed to test the quality of video coding in various scanning resolutions and modes.

But I think if we got jq963152 the masters for the hobbit, compared 720p and 1080i, and found 1080i won, he'd find some reason not to believe the result.

Cheers,
David.

[wonkey_monkey]

By the way:

Some of you were discussing 540 lines (1080i) vs. 720 lines (720p).

I, for one, am most certainly not discussing "540 lines (1080i)" - that is your (for want of a better word) prejudice and vastly, possibly disingenuously, oversimplifies the situation.

"The Hobbit" is 2.35:1.

Therefore it would even be roughly only 408 lines (1080i) vs. 544 lines (720p)... :eek:

No, it wouldn't be "only 408 lines." Even the bad old CRT days, without modern deinterlacers, it would have been considered equivalent (in terms of perceived quality) to something like 750 lines. On a still or slow-moving scene 1080i is practically indistinguishable from 1080p. The only time you could claim that it is only 408 lines (for a 2.35:1 frame) is the very very specific case where the picture is moving vertically by exactly an odd number of pixels per field. And if that number of pixels is high (by which I mean around 5 or above) motion blur will probably have already swamped any detail such that it was never there to be lost in the first place.

Besides of all of which, 1920*540>1280*720 :p

But I think if we got jq963152 the masters for the hobbit, compared 720p and 1080i, and found 1080i won, he'd find some reason not to believe the result.

That gives me an idea...

David

[dukey]

Even the bad old CRT days, without modern deinterlacers,

CRT screens don't need deinterlacers, since they are naturally interlaced displays.

[MokrySedeS]

So, how could they have done that (producing 24 fps version from 48 fps footage)?

Converting 48fps to 24fps was done simply by discarding every other frame, a process called ‘disentangling’
http://www.definitionmagazine.com/journal/2013/2/19/winning-the-frame-game.html

[SeeMoreDigital]

So, how could they have done that (producing 24 fps version from 48 fps footage)?
The easiest method is to drop every alternate frame...

Sufficed to say, if there was a 48.000 fps sample available, it would be possible to do some comparisons.

EDIT: Looks like MokrySedeS beat me too it. And has provided actual confirmation from Peter Jackson himself. Glad that's sorted ;)

[wonkey_monkey]

I read somewhere that they used a 270 degree shutter equivalent - so exposure for 1/64s in each 1/48s frame. By discarding frames you'll end up with 24fps with a 135 degree shutter equivalent. I've always used 180 degrees when I've needed motion blur, so this makes a reasonable compromise - slightly more blur at 48fps, slightly less at 24fps.

Clearly the only practical solution to this and all other standards conversions problems is for everyone to shoot at 1200fps.

[2Bdecided]

120fps, 150fps, 300fps and 600fps are possible candidates for the future, with obvious integer relationships to some/all existing formats. They already exist for slo-mo.

Cheers,
David.

[SeeMoreDigital]

I've just thought of a very good reason why a 48fps (48Hz) version of the Hobbit wont be happening any time soon.... Such a frequency does not have a supporting HDMI EDID code ;)

[colinhunt]

So, what about that 1080p48 option mentioned by "Sagittaire"
I'm pretty sure it's nowhere near as simple as Sagittaire wants to believe.

[nevcairiel]

24p@3D has 24 time slots, 2 frames each, 48p@2D has 48 different time slots, so i doubt its that easy.

[SeeMoreDigital]

Something just came to mind...:

The AVCHD 2.0 specifications.

:)

See the following thread for example:

AVCHD Spec Version 2.0 Updated 1080/60p , 50p (http://forum.doom9.org/showthread.php?t=161828)

If only Blu-ray Disc specifications would support H.264 High Profile @ Level 4.2...

Well, "Sagittaire" mentioned/suggested that all Blu-ray 3D stand alone players basically should be able to support Level 4.2 (1080p with up to 60 fps).

How true is that?
Hmmm.... Although 1080p50 has a supporting HDMI EDID code. The same can't be said for 1080p60

[kolak]

As long as it's not in the official BD spec NO ONE will replicate such a disc- purely theoretical discussion.

Most players have chips which can play files way out of BD spec, but at this point it means nothing (read line above)

[SeeMoreDigital]

What :confused::rolleyes:?

Of course HDMI supports 1080p50 and 1080p60...

It's supported since HDMI 1.0 even...HDMI may support it but I'm referring to a supporting EDID code. For which there is none for 1080p60.

See here: http://en.wikipedia.org/wiki/Extended_display_identification_data#CEA_EDID_Timing_Extension_Version_3_data_format

[nevcairiel]

And what about code 16?
16 1080p 16:9 1920x1080p @ 59.94/60Hz

[SeeMoreDigital]

And what about code 16?
16 1080p 16:9 1920x1080p @ 59.94/60Hz
Bugger... I've been starring at that list for too long and got to the stage where "you can't see the wood for the trees" :scared:

Can you double check that a 1080p 48Hz code is or is not their too?

[zambelli]

Also, virtually no HDTV today accepts 1080p47.952 or 1080p48 via HDMI (unfortunately).

Which is exactly how we got started on the whole hypothetical "How would you encode 48p in a Blu-ray/HDMI compliant video" topic. :)

[DoctorM]

Hopefully not being redundant...

I'm sort of on board with the idea that 2D generated by MVC would work just fine. The downside is you would have to require an ACTIVE 3D setup but not wear glasses. Passive 3D would be out in the cold since that works by halving height rather than doubling the frame rate.

1080p24 in this manenr would essential generate 1080p48 for any one not wearing 3D glasses. Without a parallax between the normally left and right images, it would be seamless 48p to the viewer. (Put on 3D glasses if you'd rather watch it at 24fps).

That said, I can't see it ever being released that way since it has a specific hardware requirement and requires explaining the need to have a 3D setup but it's not 3D and you don't want to put on the glasses. I can't see how anything could go wrong there.

What I didn't see proposed on this thread is another reasonable method: 1440x1080x59.94i (yes that's official BD standard).

We're talking 2:3 pulldown (or would it be called 4:6?) In theory, there might be HDTVs that could handle IVTC'ing to 48fps (with a firmware update?) since it's based on the traditional NTSC telecine pattern.

On the flip side, if you wanted 3D, the only real answer is 720p with a framerate conversion from 48 to 59.94 for compatibility.

I have a hard time imaging that ending up particularly smooth. While this would probably be the least acceptable to us, it would probably be the most acceptable to the studios.

[TheSkiller]

Maybe because someone mentioned 1920x1080 50i/25 would be too hard to compress (I don't think so, considering how much space there is).

[Sparktank]

I've noticed while watching the 2D Blu-ray with my Pulfrich glasses, that it comes in much more clearer and crisper compared to movies filmed in 24fps.
I remember going to the theaters to see it again with a friend, but only in 24fps 3D and it wasn't that bad either.
I still preferred the 48fps and, quite honestly, miss it.
It's kind of not the same watching the movie when I have that distinct memory of seeing 48fps for the first time.

[DoctorM]

Why on earth do you suggest 1440x1080i, when 1920x1080i is in the official Blu-ray Disc specification as well?

Simple:
1920x1080x29.97i, 25i (16:9)
1920x1080x24p, 23.976p (16:9)
1440x1080x59.94i, 50i (16:9)
1440x1080x24p, 23.976p (16:9)

How are you going to hit 48 fps with 29.97i?

I supposed if you assumed 48 fields.
48 frames -(resize)> 48 fields -(rate change)> 59.94 fields -(alternate frames half line up/down shift & weave)> interlaced 29.97 frames

1440x1080 allows 59.94 fps without having to drop to 720p. It'd be a lot like 23.976 to 29.97 pulldown. Little resolution is lost.

1080p29.97 would be more like 1920x540 using the method above.

With 1440x1080 you'd have more of the original frames and it IS part of the BD specifications even if it isn't common.

http://en.wikipedia.org/wiki/Blu-ray_Disc#Video

You don't have to hope that NTSC players support 25i.

It's probably the best option, doesn't require special hardware or knowledge, and it is 2D which would probably have broader sales than a 3D version.

I also maintain if it became popular (I'm thinking 3 Hobbits and possible 3 more Avatars to start), TV manufacturers might be able to support IVTC to 48p like the ones that properly play 24p content from 29,97i sources.

[Pomegranate]

DoctorM, 1440x1080 can only go up to 59.94 fields per second for BD, not 59.94 frames per second. So, it's the same problem as 1920x1080.

[DoctorM]

29.97i and 59.94i usually means the same.

Both terms usually are being used for 59.94 interlaced fields per second.

The Blu-ray Disc specification supports 1440x1080i59.94 and 1920x1080i59.94.

Except when it doesn't. I've never seen any one use the fields number when naming the interlace format.

I've seen conflicting info, but there are sources claiming 1440x1080 has 59.94 FRAMES (119.88 fields) whereas 1920x1080 has 29.97 frames with 59.94 fields. Maybe that's from people that can't distinguish the difference between fields 59.94i and frames 29.97i.

I'm not sure which sources to believe and not swearing to either. I'm just speaking based on the what I've seen.

Wrong. 1440x1080 does not allow 59.94 progressive frames per second like 1280x720 does. 1440x1080 instead allows for 59.94 interlaced fields per second.

The only thing wrong is your interpretation of what I've said. I never said 1440x1080 was progressive (although it can be at 24p), but I did say frames. Like I said, I've seen conflicting info.

It would make sense if 1440 allowed 119.88 fields since 720p supports 59.94 progressive frames.

The format seems to be a compromise between true 1080i and 720p. You'd think the higher frame rate would be what it's gaining from the trade off in resolution.

Again, it absolutely does not make sense to use 1440x1080i59.94 when 1920x1080i59.94 is also part of the Blu-ray Disc specification. My point: 1920x1080i59.94 is better than 1440x1080i59.94 because it simply is higher resolution and it's 16:9 instead of 4:3 (no anamorphic encoding needed with 1920x1080i59.94).

It does make sense IF, and I stress IF, the 119.88 number is legit. You would basically have a doubled frame rate equivalent to NTSC's old 480i29.97 (with it's 59.94 fields).

3:2 pulldown would give you a reversibly telecined film with each of the original 48 frames presented at full height (more or less).

Can anyone clarify if 1440x1080 can hit 59.94 interlaced frames/sec versus 1920x1080's 29.97 interlaced frames/sec?

[nevcairiel]

Can anyone clarify if 1440x1080 can hit 59.94 interlaced frames/sec versus 1920x1080's 29.97 interlaced frames/sec?

Both formats have 29.97 interlaced frames or 59.94 fields. One is for 4:3 and one for 16:9 content. There is no difference in frame rate.

Technically, writing 59.94i is more correct then 29.97i, but both are used for the same thing usually. One refers to the number of encoded frames, and the other one refers to the number of frames after decoding and deinterlacing (or the number of fields, if you want)
Its confusing, but people can never agree on simply using only one. It'll get worse if we actually get 120 fps content one day, which may come with 60 interlaced frames, or 120 fields, giving 59.94i a double meaning. But today, such content does not exist.

PS:
You can get an overview which formats are valid on BD here:
http://forum.doom9.org/showthread.php?t=154533

[DoctorM]

Thanks for the clarification nevcairiel.

[TheSkiller]

That's why I'm in for the terminology:

[encoded frames]i/[encoded fields]

Example: 29.97i/59.94
No doubts.


Both formats have 29.97 interlaced frames or 59.94 fields. One is for 4:3 and one for 16:9 content
No, both are for 16:9. 1440x1080 is anamorphic, it's a 1920x1080 video resized to 1440x1080 to save bandwidth.