Path: utzoo!utgpu!news-server.csri.toronto.edu!neat.cs.toronto.edu!mgreen
Newsgroups: comp.graphics
From: mgreen@cs.toronto.edu (Marc Green)
Subject: re: Psycho Graphics
Message-ID: <91Feb14.155854est.8239@neat.cs.toronto.edu>
Organization: Department of Computer Science, University of Toronto
Date: 14 Feb 91 20:59:15 GMT
Lines: 70


>From: uselton@nas.nasa.gov (Samuel P. Uselton)

>This statement I take issue with.  If we can generate images from mathematical
>models by use of a computer which are identical to images of real objects as
>recorded by a camera and played back appropriately, we have accomplished
>several things.  We have experimentally verified our understanding of the
>physics producing the image.  

True. You can produce the desired image. But thay says little about
how the image will be perceived. The question is not whether physics
can be be used to create a particular image. The question is whether
physics can tell you what image you should create in the first place.
And the answer is that physics alone cannot. The eye is not a piece of
film which simply records images.

>We have a means for producing images of objects
>that may be difficult or impossible to actually photograph.  We have a means
>to present the stimulus to the observer in EXACTLY the same way.  Do you
>claim that the perception of images will differ (for a particular observer)
>even if the stimulus is the same?  If the observer has no information about
>whether the image is "real"?

It depends on what you mean by stimulus. A patch of light will
certainly look very different at different times depending on
background, adaptation, etc. For example, simultaneous brightness
contrast. Then there are numerous reversible and ambiguous figures.
There are moving objects which sometime look stationary and stationary
objects which sometime appear to move. The list is endless.

>Whether something is unnecessary depends on the application.  They may be 
>unnecessary FOR WHAT YOU DO and still be a reasonable thing for others to
>work on.

I may not have made my point clear about camera models. The point of
camera models is to make images which look like they were generated by
a movie camera. Why? because the blurred images in movies make motion
appear smoother. Instead of worrying about the rather complicated
calculations necessary to create the camera model, why not try to
_directly_ taylor the computer images to the visual system. Why have
the camera as a middleman in the model? If blurring images make motion
smoother, then find out why an use that information in creating more
relaistic images. The goal, after all, is more realistic images, not
the achievement of some mathematical fidelity among different image
media.

If you know the properties of the visual system, you create images
which are a good match. This results in both better and cheaper
computer graphics. For example, the psychophysical literature on
apparent motion reveals a lot of tricks that could be used to reduce
the rate at which animated scenes need be updated.

Marc Green