[Tig] 4K digital intermediates using video technology

Bob Friesenhahn bfriesen
Fri Jun 24 18:26:40 BST 2005

On Fri, 24 Jun 2005, Adrian Thomas wrote:
> Well that's just it - the whole basis of digital signalling relies on 
> establishing a set of parameters that define your signal within USEFUL 
> parameters and then trying to preserve that finite information as well as 
> possible within those parameters. Just as CD Audio PCM was fixed at 16 bit 
> 44.1Khz after exhaustive blind testing, we would hope that Sony et al would 
> define standards like HDCAMSR after needs rather than technical

You pose an excellent example with CD Audio PCM.  In the '80s this was 
thought to be good enough to support the future, yet many audiophiles 
found that (other than clicks/pops) CD Audio did not sound nearly as 
good as their LPs.  CDs sounded "lifeless" as compared with LPs.  So 
now we have completing SACD and DVD-Audio which solve this problem by 
using much higher sampling quality and more bits.  CD Audio PCM is 
clearly on the decline (not necessarily due to SACD and DVD-Audio).

Another example from the 80's and early '90s is that quite a bit of 
video programming was done using analog or otherwise lossy methods 
methods.  Now we find that the quality of these efforts is "junk" and 
unmarketable in the future as re-runs because it looks so bad.  Older 
features were based on film so they can be re-captured using modern 
methods, and current features are done on film or shot directly in HD 
so they are assured a longer marketable lifespan.

> feasibilities. Is 4K based on a need or a feasibility? Why not push it to 8K? 
> Could you simply scan at 2K, do a bicubic scale up on the data and THEN JPEG 
> encode? At what point is the high frequency information so attenuated that it 
> simply makes no useful difference going further? Obviously, ANY improvement 
> in MTF at the scanning stage will improve system MTF, but what are the 
> ballpark figures for MTF50 improvement of - say - 4K vs 3K scanning? Also, if 
> it IS acceptable to transform directly to YCrCb should we be doing full band 
> RGB scanning at all, or would a Bayer-type approach be more practical, with 
> the JPEG conversion carried out prior to the creation of any RGB data?

The concern is that if corners are cut today that history will repeat 
itself and the resulting quality (while acceptable for today's 
standards) will not meet the quality of tomorrow's standards so the 
work won't earn as much money as it should.

In a couple of years, the common environment for home TV viewing will 
be 1080i HD.  It won't stop there.  Very quickly the bar will be 
raised to 1080P, and within ten years or so we will see 4K displays in 
people's homes (technology has already been demoed).  As has occured 
many times before, traditional cinema will need to compete by 
continuing to raise the quality bar so it can continue to exist.

> Sorry if all this is covered in Peter's paper, I haven't read it!

It is still available on the Cintel site.

Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/

More information about the Tig mailing list