P3D Re: depth perception cues

[roberts@xxxxxxxxxxxxxxxxx (John W Roberts)]

>Date: Tue, 9 Mar 1999 22:25:20 -0700
>From: Peter Abrahams <telscope@xxxxxxxxxx>
>Subject: P3D depth perception cues

[John R]
>>convergence is a "minor" cue for depth perception. ... implies that the
>>ability of humans to judge
>>absolute depth by binocular stereo in the absense of other visual references
>>is not very good.

>Binocular stereo is based on the disparity between what the right & left
>eyes see of the object, and convergence is sometimes a minor assist.  What
>am I missing here?

Not wanting to get into another "definition war" at this point :-),
I think what people tend to mean by "binocular stereo" is the creation
(from two-eye cues) of a three-dimensional model in the brain, regardless
of the issues of whether the scaling of that model is correct, or whether
this model is based on actual 3D input. By "ability to judge absolute
depth", I mean the ability to accurately estimate how many feet or meters
a visible object is from the observer. If I'm looking at a single "point"
source against a featureless background, then basically there *is* no
disparity between the left and right views, because the eyes are converged
on the object. Under such circumstances, if the situation is such that
chromatic and accommodation cues do not play a significant role, then
convergence becomes the dominant depth cue. So I do not think convergence
is *always* in the role of being just a minor assist.

More generally, even when looking at general scenes, disparity between left
and right views will always be partly a function of the degree of convergence
of the eyes. If convergence is a minor factor under those circumstances,
that implies that humans judge distance to objects much more by reference
to objects of known size (and related cues) than by direct stereo perception
of distance. That may be fine for many "natural" scenes, but implies that
there will be particular problems in situations without familiar objects
of known size, for example working outside in outer space, and a need
for training and deliberate introduction of additional cues to address this
difficulty.

>>I can think of a possible way to test it - when viewing a monochromatic
>>"point" light source (of adjustable brightness) in otherwise total darkness, 
>>sufficiently far away that accommodation doesn't provide a good cue, and
>>with 
>>the test subject not allowed to move their head, then convergence should be 
>>just about the only cue, and the accuracy in judging distance can be
>>evaluated.

>If it's far enough that accomodation isn't a cue, then convergence will be
>at its far limits as well (I speculate). 

Could be. Of course, in bright light, (for example with a background of a
featureless sky) the scaling of accommodation depth cues will be decreased, 
and as well as the "amount" of information available from accommodation.

>Convergence & accomodation together probably have greater sensitivity than
>either one alone. Slight mismatches certainly set off alarms in the brain,
>or cause objects to appear bigger or smaller than they are.

Good point.

John R


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