Re: stereovision question (means of perceiving depth)

[P3D John W Roberts <roberts@xxxxxxxxxxxxxxxxx>]

Looks like it's about time to repost the old list...

-------------------------------------

Date: Wed, 25 Jan 95 20:06:26 EST (first posted earlier)
>From: John Roberts <roberts>
 
Primary sources of human depth perception:
 - binocular stereo: determining the distance (or relative distance) of an
     object based on the convergence of the eyes, and the disparity (if any)
     of the two views. Note that you don't have to converge on a particular
     object to get an idea of its distance (change in disparity with change
     in convergence is another source of information). (When done properly,
     the Pulfrich effect produces true binocular stereo, though it is
     produced by motion.)
 - visual cues from the image:
     - perspective
     - known shape of an object
     - known size of an object
     - hiding of an object by other objects in front of it
     - multiple-reflection lighting (including reflection from non-specular
       reflectors)
     - distance-related cues such as haze, brightness in a flash photo,
       texture gradient (variations in perceptible detail with distance), etc.
 - focal distance of the eyes (which viewers of traditional stereo photos
     must learn to decouple from convergence)
 - depth determined from motion of the point of view (head or camera)
     examples:
       - lateral motion of head or camera
       - rotation of the object being viewed
       - oscillation of the camera (various patented techniques)
       - motion of the point of view directly toward or away from the
           scene being viewed (rate of change in size and angle of view of
           objects in scene)
 - audio effects, and combination of visual and audio effects
    (A simple example is determining the distance of a flash of lightning,
    but the human auditory system is capable of perceiving much shorter
    distances - there's circuitry in there capable of handling time intervals
    of hundreds of microseconds, or perhaps better. Humans also use
    echolocation, usually not consciously - I don't ordinarily notice it,
    but I find that I can usually walk down a hallway with my eyes closed
    without bumping into anything, and I can't think of any other explanation.)
 - kinesthetic senses: your brain knows where in space the articulated parts
    of your body are.
 - Weird stuff like sensing of thermal infrared (The temperature sensitivity
    of human skin is actually about the same as that of the sensors of pit
    vipers, though handicapped by being attached to warm flesh. Humans can't
    image in thermal infrared like pit vipers, but they can detect the
    proximity of hot or cold objects.)

Note that many of these methods of depth perception are very peripheral
to stereo photography, but applications such as advanced "virtual reality"
will require the clever combination of multiple depth perception techniques
for the most effective creation of realism.

-------------------------------------

And now for a *new* comment: in the "real world", one would
expect all the cues that are operative in a particular case to more or
less corroborate one another. There is no stereo photography/viewing system
or computer imaging system I am aware of that causes *all* the cues to be
consistent with one another. Thus when viewing a photographic or computer-
generated stereo, there will be some sort of conflict among the cues, and
the way in which this conflict is handled is a major factor in determining
how understandable and enjoyable the stereo will be. While it is distinctly
possible that no single interpretation will be produced, in the more
successful stereos one can generally assume that the weight of evidence
will favor a particular interpretation of the depth cues. But while the
basic ability to utilize binocular stereo vision may be developed only
during a certain period in infancy, refinements in this and other skills
can be learned throughout life, and the perception of a stereo photo or
computer image is largely a learned ability. The relative weighting assigned
to various cues therefore varies from person to person, and over time in an
individual. (For example the "orthostereo" people place much greater weight
on certain cues than most other people do.)

This disparity in the use and weighting of depth cues among individuals
is probably the basis for much of the argument on this list on what
constitutes a "good" stereo image. Experienced viewers should be aware
that their training may allow them to disregard conflicting cues that bother
newcomers, or may cause them to notice and be annoyed by problems that
newcomers overlook. Whether the primary objective of training should be to
get the greatest possible enjoyment out of the widest possible range of
stereo images, or to be irritated to the greatest extent possible by any
deviation from "perfection" (or somewhere in between) is a matter for 
philosophical debate. My personal preference would be to enjoy a fairly
wide range of works, while not failing to see how they could be improved.

John Roberts


------------------------------