[opendtv] Re: (No Date: Tue, 28 Sep 2004 10:30:37 -0400

  • From: Tom Barry <trbarry@xxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Tue, 09 Nov 2004 19:53:43 -0500

inline ...

Craig Birkmaier wrote:

> At 12:38 PM -0500 11/8/04, John Shutt wrote:
> 
>>And I have argued that this is how MPEG 2 should compress an image.  DCT the
>>edges first, lowering the high frequency information from the edges towards
>>the middle as necessary to meet your target compression rate.

...
> 
> Three problems.
> 
> First, if we are talking about SDTV, every sample is critical. The 
> typical display covers an area that is roughly equivalent of foveal 
> vision. Thus there is no need for directed eye movement, as the 
> entire display will have the same apparent sharpness, without the 
> need to track moving objects.
> 
Somewhat true for SDTV, but see below

> Second, for HDTV - or any form of moving imagery on a big screen - 
> you cannot control what the viewer eye-tracks. If they don;t look at 
> what has encoded to be sharp, they will see compression artifacts. 
> And critical info may very well come in from the edge of the screen.
>
You can somewhat control it simply by placing it in the center and also 
by having it in focus.  This is often done now with lighting and depth 
of field keeping the attention on the desired parts.  It would be even 
more so if the center of the screen appeared to have more detail, 
assummming it was the center you wanted viewers to see.  Often the edges 
of the screen are just a "safe area" anyway and slight softness of 
detail would tend to direct viewers away from this area.

> Third, the DCT does not provide the tools to soften the samples 
> within a block - all it can do is quantize away coefficients making 
> them more alike. What you would see is blocking artifacts and 
> quantization errors near the edge of the screen, not "soft focus." 
> You COULD, however,  pre-filter the imagery to be softer near the 
> edges, then compress it.
>
I think the idea would be instead to use non-linear scaling like the 
sample I prevously posted, not a custom DCT.  That scaling is reversed 
at display time to create a linear image but with a non-linear detail 
density or MTF.

- Tom

> 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.
> 
> 
 
 
----------------------------------------------------------------------
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:

  1. Home
  2. opendtv
  3. Archive
  4. 11-2004