[opendtv] Re: Half Truths - Was More 1080p@60
- From: Craig Birkmaier <craig@xxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Thu, 6 Dec 2007 09:14:51 -0500
At 10:03 AM +0100 12/6/07, Jeroen Stessen wrote:
Hello,
Craig wrote:
Display Oversampling.
I think that on the next CES we are going to show you something
interesting, like temporal up-conversion (50 to 100 fps) on LCD.
That really makes a big impact, especially on pans. 50-60 fps
is so passé, you really want 100-120 fps... ;-) Stay tuned.
PS, 90+% of the credit is for our friends at NXP Semiconductors.
They did a great job on HD Natural Motion in the new PNX5100.
http://www.google.com/search?hl=en&q=PNX5100
Thanks Jeroen!
It is unfortunate that there is so little
understanding about the decoupling of
acquisition, emission and display, and how to
optimize each to deliver the highest quality to
the viewer. We placed a great deal of emphasis on
this in the 1992 SMPTE/EBU Task Force Report on
Digital Image Architecture.
As we have learned with the evolution of computer
display systems, there is a great deal that can
be done to enhance the presentation of different
types of imagery using the image processing
capabilities of the display processor. In this, I
believe that Philips has been a leader, starting
with the Tri-media processors back in the '90s.
Having spent much time working with HDTV
acquisition and display systems, we learned early
on how to acquire, encode and display imagery to
maximum benefit. I fully understand why the
100/120 fps numbers are now emerging - this is
the easiest way to deal with the 50/59.94Hz
heritage of analog legacy formats, and for the
120 fps crowd, we get an integer multiple of the
important 24 fps source library (actually 23.97).
While I am certain that your neww developments
will look very good, it is unfortunate that we
did not take advantage of the DTV transition to
move to a more friendly/useful set of frame rates
that are a better match for the human visual
system. The task force recommended a family of
frame rates with integer relationships; the most
important members of this family being 24/36/72
fps. Itis important to note that for bandwidth
conservation this family could also support 12
and 18 fps. As you are well aware, the flicker
fusion rate for the average human - where we
start to see continuous motion - is about 18 fps.
But at low frame rates like 18 or 24, we need
careful control of the viewing environment to
prevent the perception of flicker (i.e. a
darkened movie theater). By the time we get to 36
fps we get very good motion continuity, although
this is not adequate for rapid motion like
sports. For sports, 72 fps would have been a good
target, as it provides excellent continuity in
rapid motion, better instant replays, and the
frame rate is high enough to prevent the
perception of flicker even in brightly lit
environments - there were good reasons why Europe
established 72 fps as the floor for computer
display refresh rates!
It is interesting to note that 30 fps has never
taken off as an acquisition frame rate. I believe
that there are several reasons for this:
1. First, it is not that different from 24 fps,
which IS a widely used acquisition frame rate. We
have seen a major shift to cameras that can
acquire at both 24 and 30 fps, but only 24 fps
has seen widespread usage.
2. 30 fps still has motion discontinuity issues like 24 fps.
If we have a family with 24/36/72 there is better
separation between the rates. 36 fps become
desirable for application like talking head shows
- it could also be used for virtually everything
that Hollywood produces, although it is unlikely
they will abandon 24 fps.
Underlying all of this, however, is a very
important reality. To implement features such as
motion compensated frame prediction, which is
needed to gain the full benefit of 100/120 fps
display, you need high quality samples. It is
impossible to do a good job with this if you are
starting with samples that have severe
compression artifacts including blocking and the
sample distortions introduced by excessive
quantization. Just look at how most HDTV capable
displays deal with legacy interlaced video
sources - they make crappy video look soft to
hide all of the artifacts of the digitizing and
deinterlacing processes needed for progressive
displays. And the image pre-processing that takes
place during emission encoding removes critical
detail that your image processing algorithms need
to do a good job.
Alas, it may be another decade before the
television industry figures out that they are
using the wrong family of frame rates, and that
it is MORE IMPORTANT to deliver high quality
samples than a high quantity of impaired samples.
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:
