Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!ginosko!uakari.primate.wisc.edu!ames!ames.arc.nasa.gov!watson
From: watson@ames.arc.nasa.gov (John S. Watson)
Newsgroups: comp.graphics
Subject: Re: Smart Pixels?
Message-ID: <33437@ames.arc.nasa.gov>
Date: 12 Oct 89 05:17:37 GMT
References: <207400033@s.cs.uiuc.edu>
Sender: usenet@ames.arc.nasa.gov
Organization: NASA - Ames Research Center
Lines: 27

In article <207400033@s.cs.uiuc.edu> mccaugh@s.cs.uiuc.edu writes:
>
> Does anyone know of a system with "smart" pixels? 

Once upon a time I wrote a ray tracer in which the pixels
used hueristics to determine thier sampling rate.
Since the  reason for doing it was to speed things up, 
the hueristic had to be simpler than casting a ray( s, if sub-sampling).  
I used difference in previous pixels values, with a little randomness 
tossed in.  So pixels changing quickly were sampled every frames, 
while pixels that were hardly ever changing were sampled only once every 
10 frames.
The results: much faster, but with a graininess on the edges of moving objects.
I needed to make the pixels more aware of what was happening with its neighbors.
Never got around to doing that.

Another problem is big pictures have lots of pixels ... 512x512 = 0.25 million.
To be smart you must have a memory.

To save memory, I combined the above with an Area-of-Interest/Variable Acuity
(AOIVA) Ray Tracer.  

Hope this helps, 
John S. Watson, Civil Servant from Hell        ARPA: watson@ames.arc.nasa.gov 
NASA Ames Research Center                      UUCP:  ...!ames!watson
Any opinions expressed herein are, like, solely the responsibility of the, like,
author and do not, like, represent the opinions of NASA or the U.S. Government.