Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!elroy.jpl.nasa.gov!jarthur!ucivax!honig
From: honig@ics.uci.edu (David Honig)
Newsgroups: comp.graphics
Subject: Re: PERCEPTION (was: Turing test, psychographics, ..)
Message-ID: <27D52886.15091@ics.uci.edu>
Date: 6 Mar 91 17:36:06 GMT
References: <4996@cui.unige.ch>
Reply-To: honig@ics.uci.edu (David Honig)
Organization: UC Irvine Department of ICS
Lines: 42

In article <4996@cui.unige.ch> pun@cui.unige.ch (PUN Thierry) writes:
>As a particular exemple, one of the problem we are working on is how a 
>focus of attention is selected, ie. why and how one looks first at some 
>particular points in a scene. Clearly, such a topic is interesting for a 
>computer vision system. Our feeling is that it is also of interest for 
>image synthesis: which will be the part in a synthetized picture that will 
>be scrutinized first?

At the recent SPIE conf in San Jose someone (sorry, don't have my 
program with me) reported a study where they tracked the eye motions of
people watching canadian TV.  Since you only see well in the fovea, 
if you could send high-quality info only where, say, 99% of all people
watched, you could save megabandwidth.  BUt you need to predict this
(set of) location; either with humans (tedious) or machines (unknown how to
do this). 

>More generally, it would be very interesting to start a discussion on
>how perceptual studies can be relevant to imaging sciences. This concerns
>computer graphics as well as computer vision. The Eurographics Working
>Group on the Relations between image analysis and image synthesis intends
>focusing on issues of perception and it is open to interested researchers.
>If there is enough interest, a mailing list could be started for exchanging
>information. Interested people should contact one of the two persons
>mentionned below.

The whole reason that JPEG allows *you* to choose the quantization (as will 
MPEG) is so that *you* can implement what *you* think the human observer
can/needs to see.  

The point was made at this conference that there are two kinds of
redundancy you can exploit: image redundancy and the insensitivity of
the human visual system.  The latter gets at the importance of
psychophysics.

Cheers.


-- 
David A. Honig

"The fundamental principle of science, the definition almost, is this:
the sole test of the validity of any idea is experiment." -- R. P. Feynman