Re: Effects of Eye Position on Stereo Perception

[P3D Larry Berlin <lberlin@xxxxxxxxx>]

>Date: Fri, 18 Apr 1997 
>From: P3D Peter Homer writes:
...................
>This topic would seem to be relevent to this group when it concerns stereo
>perception. The eyes converge on objects they are viewing and it was quite
>a logical early view that this convergence gave information on depth by a
>kind of triangulation. But I had a photocopy from a book on psychology that
>stated that that theory and experimentatation showed that convergence of
>the eyes played little or no part in depth perception . Stereo viewing with
>stereoscopes and particularily free viewing would seem to support this .

******  IMHO this conclusion is very wrong. Triangulation is heavily used
but it is a relative thing, comparison triangulation, not a *fixed triangle
means X* kind of thing. It is operative in any stereo viewing circumstance
whether it be a stereo microscope, Red Button viewer, LCS goggles, or
free-viewed stereo pair.  Within a stereo view one's eyes are triangulating
on specific features during the eye-scanning process. It is very much a part
of depth perception regardless of their lack of ability to measure or look
for the right things.

Particularly important to keep in mind is that seeing is NOT done with the
eyes! How often scientists seem to forget such a fundamental thing. Actual
perception is based on mental constructs based on accumulated information,
some of which comes from the eyes. That's why we can turn our heads and not
have the whole world turn too. That's also why I feel that hyper stereos can
be in some circumstances more realistic than strictly ortho pairs. In
reality one's mental construct takes in more information about the shape and
relative depth of objects and scenes, than is possible to see with one's
eyes in a single instant or a camera set exactly where the eyes are located.


>.............................. Apparently his and
>later Brewsters reason for using half lenses originaly was not so much to
>diverege the lines of sight out to pairs larger than the interocular. But
>to diverge them on the other side of the lenses so that they converge on a
>point between the pairs.

********  Probably accurate for those devices. It's also the basis on which
LCS viewers or polarized projections work. The eyes do converge on the
stereo information. Such convergence is convenient and easy on the eyes.
Even with side by side pairs, one views what appears to be the center
overalpping image for the stereo view. Convergence of information.


>...................... What about those with
>large fixed eyepieces would only those people with the right eye seperation
>to see throught the inner halves be able to see a proper stereeo image?.
>With parrallel "free viewing"  we can be even more certain that our eyes
>are not convergent but we see a stereo image and as our brain has not
>evolved for viewing stereo pairs but real objects it is likely that this
>mechanism is involved in the viewing of such real objects. 

********  The mechanism isn't absolute convergence, but feature convergence.
In parallel viewing of two identical images one sees a flat reconstruction.
Each point in each picture is exactly the same distance apart (zero
stereopsis, zero triangulation differences). When viewing a stereo pair in
parallel, the various features are not equally spaced. It is this
triangulation of the smallest stereopsis that comprises the *convergence of
interest*. This is a relative convergence within the scene, not necessarily
an overall physical eye convergence. This is the same process we use in
viewing physical objects. We triangulate essential points of any object. All
stereo views have similar triangulated reconstruction, that's why they work.

This can be further understood by noticing the effect of viewing
stereograms. Even for the experienced, a new stereogram can take a few
moments to resolve into depth structures even after the eyes are in the
*right position*. The right position isn't where the whole image resides,
but only a convenient starting point. At first the similar patterns are just
that, similar. Then the realization of stereopsis and it's minute
triangulation comes into play and the patterns take on shape and depth. Once
resolved the focus of attention can travel to actual triangulated places in
the image. The image has become 3D.

Larry Berlin

Email: lberlin@xxxxxxxxx
http://www.sonic.net/~lberlin/
http://3dzine.simplenet.com/


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