[opendtv] Re: Time to give up on 1080i for football
- From: Craig Birkmaier <craig@xxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Tue, 8 Dec 2009 07:41:59 -0500
At 8:07 PM -0500 12/7/09, Albert Manfredi wrote:
I would readily agree that by some measures of efficiency, coding
progressive is more efficient than interlaced. For instance, I have
no doubt that coding 60 frames per second in progressive mode is
more efficient than coding 60 frames per second (120 fields per
second) interlaced.
I also wouldn't doubt that a heavily pre-filtered 1080/60p can fit
in the same channel as a 720/60p. But then again, the pre-filtered
1080/60p would also have no more detail than the 720/60p.
If we are really to believe that 1080/60p takes up no more channel
capacity than 720/60p, then we are saying that 480/60p should take
up no more capacity than 720/60p, and by extension, no more bit rate
than 1080/60p. Which I find hard to swallow.
Bert
You have unknowingly answered your own question here.
We are dealing with information theory here. Entropy Coding simply
means that we are removing redundant information from the source. IF
you reduce the information content to the same level for 1080P, 720P
and 480P, then in theory you "should" be able to encode all of them
at the same bit rate.
That being said, there IS additional overhead as we increase the
resolution of the source rasters that we are encoding. If we keep the
transform and block structures the same, there are obviously more
blocks to encode as the raster size increases, and there is more
overhead to keep track of all of this.
There were some early studies that suggested that a relatively large
portion of the bit stream for 1080i sources was consumed just with
the overhead of the DCT blocks and motion vectors needed by the
decoder to reconstruct the bitstream. And, since MPEG-2 uses a very
crude means of prediction, the encoder can quickly be overwhelmed
with the difference information from the predictions. This is the
main reason that the version of the SEC Championship I watched was so
poorly encoded - i.e. at the full ATSC channel bit rate the source
was already significantly impaired, but the artifacts were less
perceptible; after re-encoding at a lower bit rate the margins were
obliterated and the source fell apart any time there was a little
stress.
Now back to the basic question here. What is different about 1080P,
720P and 480P native formats?
The number of samples and the accuracy of those samples. As we
increase spatial resolution we give the encoder more information, but
we also reduce entropy - i.e. the accuracy of the sample relative to
what the camera actually "sees." As we increase frame rate, we gain
two benefits:
1. We can reduce the filtering needed to provide continuous motion -
i.e. we need motion blur to make up for temporal undersampling, which
is one of the major reasons that 24P source is NOT SHARP in moving
areas of the image. At higher frame rates the samples are sharper and
easier for the prediction routines to use to make more accurate
predictions.
2. We can use more B frames with higher frame rate source, which are
substantially more efficient than I and P frames.
Thus, while there is more information in higher frame rate source,
there is also more redundancy and it is easier to track this
redundancy, thus improving coding efficiency.
As an aside, interlace mixes spatial and temporal information, making
the task of recovering the original samples much more complicated. In
essence, interlace increases entropy, which is just the opposite of
what we want in a digital TV system.
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:
