[opendtv] Re: 20040921 Twang's Tuesday Tribune (Mark's Monday Memo)
- From: Tom Barry <trbarry@xxxxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Sun, 26 Sep 2004 13:18:32 -0400
Craig Birkmaier wrote:
> You are not missing 1440 x 1080i. The reality is that this is all you
> are getting, even if the encoder is set to 1920 x 1080i. There is no
> useful information to fill the spectra above that which can be
> reproduced with 1440 encoded samples. Few if any HD camera/recording
> systems even try to record 1920 samples per line; And then the MPEG-2
> encoder quantizes any high frequency details that might exist away.
> The only way to get 1920 samples with "filled spectra" is to
> oversample - the only source available today that can do this is film
> that is then scanned at 19920 x 1080.
Under current practices I would probably not be missing any extra detail
with 1440i. What I would be missing under 1440i would be the extra bit
overhead of coding 1920x1080i even when there is little underlying
source detail. That extra does not come for free under MPEG-2. No
matter how quantized or soft it is there is still some fixed overhead of
larger resolutions and that takes away from my bit budget to represent
the real detail that could actually be there to describe the picture.
So unless they (whoever they are) anticipate starting to use some more
detailed material 1920 just seems wasteful to me.
- Tom
> At 12:35 PM -0400 9/25/04, Tom Barry wrote:
>
>>I'd agree that, when properly done, 1024x576p could probably match
>>almost all of what is currently sent in HDTV today, especially if it was
>>really in 4:2:2 format. But, as you said, it mostly does not exist
>>here. And 720p is only 25% greater in each direction so it's probably
>>close enough to have similar economics.
>
>
> One important note here. For a progressive raster, 4:2:2 buys you
> NOTHING. 4:2:0 is absolutely appropriate for a progressive raster,
> since it is easy to subsample the color difference signals in both
> the H & V axis. Having full color resolution in the vertical and half
> in the horizontal does little if anything to enhance the image
> quality. If, however, you were to use 4:4:4 encoding, then there
> would be a significant improvement in color detail. This is quite
> feasible, especially for an improved DVD presentation. It probably is
> not that important for a terrestrial DTV broadcast system.
>
> There is a growing consensus that 720P is an attainable goal, so
> there is little reason to bother with EDTV. This may well be true,
> but it ignores the reality that MPEG has been designed to deliver
> almost ANY raster efficiently - it is the job of the local image
> processor to scale the raster to the local display resolution. So, as
> we already have seen with streaming video, formats are just a
> marketing tool for the companies that produce production gear. In the
> world of distribution, formats simply do not matter.
>
>
>>The intermediate resolution I miss would probably have been something
>>like 1280x1080i or even 1440x1080i since I think this would be closer to
>>"effectively square" pixels in a 16x9 format. But I suppose I have a
>>bias towards interlaced resolutions since most HD RPTV's like mine still
>>use them. This of course may change with new technology.
>
>
>
> You are not missing 1440 x 1080i. The reality is that this is all you
> are getting, even if the encoder is set to 1920 x 1080i. There is no
> useful information to fill the spectra above that which can be
> reproduced with 1440 encoded samples. Few if any HD camera/recording
> systems even try to record 1920 samples per line; And then the MPEG-2
> encoder quantizes any high frequency details that might exist away.
> The only way to get 1920 samples with "filled spectra" is to
> oversample - the only source available today that can do this is film
> that is then scanned at 19920 x 1080.
>
> The Digital Cinema projectors using TI's 1280 x 1024 DLP chips with
> an anamorphic lens to stretch out the raster to proper width come
> very close to the 1280 x 1080 raster you mentioned, except that this
> is progressive, not interlaced.
>
> Regards
> Craig
>
>
>
> ----------------------------------------------------------------------
> 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:
