[opendtv] Re: MPs back Ofcom stance on spectrum sale

At 12:56 PM -0400 5/24/07, Tom Barry wrote:

I guess you guys just like to argue.
Bert specified the maximum amount of information in what I assumed was a pure theoretical sense.
Perhaps, but he showed no indication of understanding the difference between information content and sample rates.
You specified practical real worlds cases where images at increasing resolution do NOT have flat MTF curves.
I described the issues that impact the actual information content and gave some examples to explain this. 



You are both right.



If you say so.
One of the images I play with is a huge panoramic still image of the Valley of Fire that I've previously turned into a moving pan at 1080p/60 and posted on this list. Something like that can have a huge amount of information due the opportunity for oversampling. Oversampling cameras will eventually be able to do this to. And I personally believe that reality has almost infinite detail if we could actually capture it properly.
I think it would be accurate to say that there is almost infinite detail in nature, however, when we sample reality into a 2D spatial plane at a given sample rate, we place finite bounds on the amount of information that can be captured. This is the basis for sampling theory.
Of course right now the higher spatial frequencies of any video tend to be quite limited due to the filtering of lenses, sensors, and whatever that tend to average adjacent values. This also used to be true for anything over 320x200 resolution too. But over the years the almost flat part of the MTF curve is lengthening to the right with new technology and demand. So we should not assume there will always continue to be little high frequency information or that we will be unable to compress it when there.
All true to some extent. But there are both theoretical and practical limits to how much information we can throw away before we begin to introduce distortions that can be seen by the human visual system. Obviously, if there is no limit on the peak bitrate that can be used to deliver a compressed image stream, we can keep these distortions to a level that cannot be detected. Unfortunately we do not have that luxury in most distribution systems, and ESPECIALLY in a mass market DTV system.
And then there is another reality that must be considered. Eventually we will reach the point where it is non-productive to deliver additional image detail. This will primarily be driven by display requirements. This is analogous to digital audio, where we are already approaching the useful limits for mass market products, with higher quality reserved for a small premium niche market.
You mentioned 320 x 200 resolution. This raises an important point. As we sample at lower and lower frequencies, each sample becomes VERY important, as it represents more information in the scene. Thus we cannot compress at higher ratios lest we throw away information that is critical to comprehension of the scene. As we increase the sampling rate we begin to capture more details that can be distorted slightly while still delivering a good quality picture. But we must still remember that removing too much information form any source leads to visible artifacts, thus we are better off resampling to a lower resolution for emission so that we can maintain delivered image quality.
Though I will agree that currently the delivery of higher frequency video information in HDTV is fairly non-existent.
YUP! It would be fairly easy today to demonstrate that resampling to 1024 x 576 from any of the HDTV formats, then compressing properly would deliver equal or better quality that what we see today.
This is where I believe Bert is having problems. He seem to think that we are already delivering good quality HDTV with MPEG-2. The truth is that even with H.264 we will still be lacking in bandwidth to handle the peak bit rate requirements. 

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: