[opendtv] Re: Time to give up on 1080i for football
- From: Kilroy Hughes <Kilroy.Hughes@xxxxxxxxxxxxx>
- To: "opendtv@xxxxxxxxxxxxx" <opendtv@xxxxxxxxxxxxx>
- Date: Fri, 11 Dec 2009 03:19:54 +0000
In the world of perpetually bitrate starved Internet streaming, the SMPTE VC-1
codec often gets the best results because it dynamically subsamples pictures
for optimum quality/bitrate balance. For instance, a 1920x1080 image might
downshift to 1440x1080, 960x1080, 960x720, etc. sampling under stress, which
reduces the potential resolution of the decoded and resampled 4:4:4 1920x1080
image (as does spatial prefiltering) but also reduces the sampling overhead to
devote more bits to picture quality (especially preserving low frequency
structure to avoid blocks, mosquitoes, etc.).
In addition, what has enable TV over wild Internet (like Hulu, Fancast,
NetFlix, etc.) is adaptive bitrate streaming, where selection is made every few
seconds between alternate streams encoded at graduated bitrates. The top end
for 1920x1080 might be 10Mbps, and the low end 1.0Mbps. You can be pretty sure
that anything except talking heads is subsampled at 1.0Mbps.
H.264 has similar ability in theory to change sample shape and structure on the
fly, but I don't think it is done much because it is more complex (and there
are probably decoders out there that didn't implement the full flexibility of
the standard, so any encoders that try to utilize the feature will be crucified
by the usual suspects that only support the decoding features they like in
their TV sets, game machines, etc.).
Most of the big studios, cablecos, telcos, CE manufacturers, IT companies, etc.
got together recently to standardize how video will be sent over the Internet.
The IT companies said 98% of it today is square sample, progressive, encoded at
the actual shape of the picture (e.g. most movies being 1.85 to 2.40 aspect
ratio, approximately none are 1280x720 or 1920x1080).
To reach consensus, we were forced to limit to 4:3 and 16:9 frames padded with
black bars, include 11/10 and 44/33 samples, interlace, etc. because CE
companies claimed to have devices that would explode or something if they saw
encoding that actually used the options in the H.264 Standard in ways that
did not honor the broadcast and disc tradition (even overscan for all those CRT
based portable media players and cell phones out there). In fact, most new CE
products with network and memory card ports are decoding square pixel,
progressive, actual frame shape encoding coming from NetFlix, etc.; but the old
engineers are stuck in the last century and won't admit it.
Same old game where the worst implementer with the biggest political machine
determines the limits of the Standard (where interoperability is the top
priority). Meanwhile, Apple is doing actual image shape, square pixel,
progressive encoding at native frame rate because it's the best technical
solution and pays off in picture quality per bitrate, ability to automatically
adapt different picture aspect ratios and resolutions to different screen
aspect ratios and resolutions, etc.
Kilroy Hughes
-----Original Message-----
From: opendtv-bounce@xxxxxxxxxxxxx [mailto:opendtv-bounce@xxxxxxxxxxxxx] On
Behalf Of Tom Barry
Sent: Thursday, December 10, 2009 7:13 AM
To: opendtv@xxxxxxxxxxxxx
Subject: [opendtv] Re: Time to give up on 1080i for football
Craig Birkmaier wrote:
> At 10:32 AM -0500 12/9/09, Hunold, Ken wrote:
>> Looks we've started up the old 720p/1080i argument again (in just a
>> slightly different form.) This "church bus" has been joined by many
>> and is racing downhill at an ever-increasing rate of speed.
>>
>> Let me try to divert the discussion into a "run-away truck" emergency
>> off-ramp, of sorts, by suggesting that 60 Hz is not a high enough
>> frame rate for football (either American or "metric") regardless of
>> how you slice the picture. To me, there is no advantage for either a
>> blurry 1280x720 frame or a blurry 1920x540 field. The minimum frame
>> rate seems to be 3x the frame or related field rate (150/180 Hz.)
>> Evidence to support this can be found in the 3x "Super Motion" system
>> from the 1980's (Super Slo-mo to some) and some of the higher frame
>> rate versions being employed on sports broadcasts today.
>
> We pushed very hard for a more versatile family of frame rates when
> the ATSC standard was developed. We even got Polaroid to build a
> 1280x720@72P camera to demonstrate the superiority of 72P over 59.94P.
> An interesting side note is that 36P actually works quite well for
> many applications, providing motion continuity that is far superior to
> 24P. More important, a 72P display did a much better job with 24P film
> source (triple shutter versus 3:2 pulldown), could present 36P with
> double shutter, and 72P native looked MUCH better with slo mo than
> 60P.
>
> As Ken suggests, temporal oversamplng has many benefits, especially
> for slo motion replays. But it is not yet feasible to deliver these
> higher frame rates to the consumer, nor is it necessary. THere are
> limits to how much information a human observer can process too. 72P
> is more than adequate for real time video, and the sharpness of
> individual frames is good enough for many slo-mo applications. But
> Super Motion also has a place if you want to watch the rotation of the
> seams on a baseball traveling at 90 MPH.
>
> Regards
> Craig
>
I kind of suspect that new codecs will evolve over time that present video much
more like video games than the current codecs that just display a linear
succession of frames. Morphing and moving objects will persist for many
display frames, rotating, accelerating, growing, shrinking, and undergoing
various changes in lighting conditions. Under that scenario we will probably
find that further increasing the display frame rate does not even appreciably
increase the bit rate at all, at least for scenes with fairly well behaved
rapidly moving but recognizable objects.
Original capture at 100's or even 1000's of frames / second will be a bit hard
but after some massive number crunching the near losslessly compressed video
won't really grow much with frame rate increases.
I recommend it for future studio masters. (Once we finally figure out how to do
it) ;-)
- Tom
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts:
