[TML] New 3d computer simulation design

[shadow at shadowgard.com]

On 3 Dec 2007 at 15:48, Timothy Little wrote:

> On Sun, Dec 02, 2007 at 07:16:25PM -0800, shadow at shadowgard.com wrote:

> > Video gets interesting. 2 hours of high quality video (DVD) uses a
> > bit over 4 gig. But you can get quite acceptable results in 1 gig.
> > Below that you start getting *serious* digitization artifacts.
> 
> You can get picture quality better than many analog TV receptions in
> about 150 MB per hour for typical TV content.  Certainly nowhere near
> DVD quality, but better than most people watched 20 years ago.
> 
> But yes, quality does decrease pretty drastically with bitrate.

I find anything below 350/hr is more messed up than I prefer to 
watch. At least in part because the losses aren't simple "fuzziness" 
but image artifacts that get really annoying.

> > 3d moving images will only be worse...
> 
> Probably not as much worse as it might seem at first.
> 
> You can nearly always only see the surfaces of objects, and so most of
> the time you only need to add a bit more data for the sides/rear of
> objects and add depth information.  Better yet, depth should compress
> even better than surface appearance, and would drastically improve
> motion estimation algorithms.  It would likely even greatly *reduce*
> data used by video formats, since failure to distinguish and re-use
> data from moving objects is by far the biggest cost in bandwidth.

By three-d, I *don't* mean "stereo imagery" like three-d movies. I 
mean the sort of image you can project into space and walk around to 
view from all sides.


--
Leonard Erickson (aka shadow)
shadow at shadowgard dot com