[opendtv] Re: 625 video quality is good enough....
- From: Craig Birkmaier <craig@xxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Tue, 19 Oct 2004 09:54:42 -0400
At 8:31 PM -0400 10/18/04, Tom Barry wrote:
>I know there is also a lot of theoretical support for the idea of an all
>progressive system. But I still cannot convince myself that the best
>sampling grid places the samples at exactly the same points at each
>moment of time. There seems to me to be an informational advantage in
>sampling slightly different points on each frame and letting either eye
>motion tracking or motion compensated interpolation gather a bit of
>extra detail. Though if I was given a choice I might not just pick
>alternating even and odd lines.
You are confusing sampling and emission encoding, which are
completely decoupled in a digital system.
As we have discussed many times, the key to proper sampling is
OVERSAMPLING. As you suggest, one way to do this is to change the
sampling location. Another is to increase the sampling density such
that multiple samples contribute to the optimized samples that are
encoded for emission.
What is well documented is that undersampling does not work well as
it creates gaps in the spectra that must be guessed at when
processing the imagery. Interlace is an undersampling technique that
trades spatial detail for temporal detail. It present REAL problems
for entropy coding techniques, since it increases image entropy.
>But I do believe that some of the problem is that neither the codecs nor
>most fixed pixel displays deal with interlace very well. Those problems
>may be the limitations of current software and not necessarily a show
>stopper.
The reasons for your observation are obvious. Interlace worked
adequately for scanning displays that are properly set up to overlap
the fields. The vertical resolution is cut by half. The human visual
system does the reconstruction, by integrating a moving spot of light
so that we see moving images.
When you illuminate the entire display for the full field/frame
period this nifty little "trick" falls apart. You must create the
missing information in the spectra to fill in the missing lines. This
is an imprecise process at best. Likewise, an MPEG-2 encoder must
deal with these spectral gaps. There are some tools for encoding
video fields that help, but they cost you in encoding efficiency.
What is worse, is that it is impossible to properly scale the color
difference signals to achieve the 4:2:0 space that MPEG-2 uses for
emission encoding; you must combine two video fields then scale,
introducing a wide range of color artifacts, or you must de-interlace
prior to scaling, a feature not supported by most encoders. On the
other hand, it is trivially easy to scal a progressive raster to
4:2:0.
The only displays I have seen that deals adequately with interlaced
HD are the ALIS plasma panels. In essence these panels are designed
to simulate the way an interlaced display works, shifting each field
up or down appropriately. IMHO, even these displays do not provide
the quality of a good progressive display.
Once again I will state the obvious; we don;t need an analog
compression technique that undersamples TV images.
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.
Other related posts:
