[opendtv] Re: AVC realistic bitrates?
- From: "Don Munsil" <don@xxxxxxxxxx>
- To: <opendtv@xxxxxxxxxxxxx>
- Date: Thu, 09 Jun 2005 10:45:32 -0700
From: <dgrimes@xxxxxxxxxxxxxxxxx>
> First, we have been arguing the quality of picture from two different
> perspectives: objective and subjective. We cannot subjectively argue
> whether there is a 10:1 or 3:10 or 2:1 ratio in b/s for the same quality.
> We will all judge differently. However, on an objective judgement, we can
> compare the differences mathematically. If PSNR is the criteria, we can
> easily determine the difference, and thus ratio, of the two in terms of
> bits per second for the same quality. If we use different criteria, the
> ratios will naturally be different.
I agree with the substance of your post, that "bits-for-equivalent-quality
ratio" varies at different rates, but I think PSNR is not a great measure of
"quality." It's exceptionally easy to increase PSNR and lower perceived
quality, or vice-versa. As a broad-brush "does this suck or not"
measurement, PSNR is better than nothing. But for fine evaluation, it
doesn't work well at all. For that, the best measurement found so far is
trained observation by judges.
Here are a few interesting papers with proposed quality measurements that
take into account perceptual characteristics:
http://live.ece.utexas.edu/publications/2002/zw_icme2002_videoqual.pdf
http://www-ise.stanford.edu/class/ee392j/projects/projects/xiao_report.pdf
http://wes.feec.vutbr.cz/UREL/RADIOENG/sample04/pg-48-50.pdf
Don
----------------------------------------------------------------------
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:
