Convergence as a cue to depth perception

[P3D John W Roberts <roberts@xxxxxxxxxxxxxxxxx>]

>Date: Tue, 17 Jun 1997 10:44:40 -0500
>From: P3D John Bercovitz <bercov@xxxxxxxxxxx>
>Subject: Re: How to fool people with 3d

>I am surprised to hear from John R that vergence is such a good 
>depth clue when papers and lore of researchers seem to indicate 
>otherwise.  Perhaps we need to do our own experiments here on P3D?  
>I think vergence has always been considered a fair depth clue out 
>to a few feet but not beyond.  I would love to hear of proof that 
>this is not so.  Conflict of vergence and other stereopsis clues
>occurs when film chips are badly spaced.  If they are very badly
>spaced, it's obvious because things look too big or too small.

Two of the sources I like to use are not available at the moment.
Ferwerda's book discusses this issue in chapter 21 ("About viewing
stereoscopically"). This book refers to a number of experiments,
and claims that cues due to convergence are only good for convergence
of 2 degrees or more, as compared to mean sensitivity to retinal
disparity of 5.6 seconds of arc. But I'm personally skeptical of that
2 degree number. The experiment described involved test subjects being
shown "a light" in a dark room for a short time - estimates of the distance
varied from 3 to 6 meters. The Ferwerda book doesn't explain why seeing
the light for "a short time" was considered desirable - I would think
that if you want to test feedback from convergence, you would want to
allow plenty of time for the convergence to be performed.

The overall tone of this chapter and the description of the experiments
seems to be an attempt to disprove a contention that convergence is the
*only* cue to depth, or that convergence is *required* for depth 
perception - so I don't know how valuable it is in figuring out
the cooperative roles of convergence and retinal disparity in depth
perception. The extreme difficulty in avoiding 2D cues over larger distances
adds to the difficulty of devising experiments to separate the perception
mechanisms of convergence and disparity from one another. 

I agree with John B that some experiments might be valuable. But first,
what do we mean by "depth"? Is it the sense of knowing that one object
is closer than another, or the sense of knowing how far it is to an
object (or some combination of the two)?

If retinal disparity is considered to be the primary cue to distance (a
premise of which I remain somewhat skeptical), then what is the model for 
judging the distance of an object that's 100 feet away? Do you use retinal
disparity to note that it's 10 feet further than an object 90 feet away,
which in turn is 10 feet further than an object 80 feet away, and so on,
building up a whole framework of relative distances? If that's the case,
then what's the "absolute" upon which this framework is constructed? Does
there have to be some object you can see that's close enough for convergence
to play a major role, in order to be able to judge the distance to objects
that are further away?

If you see an object (say 100-300 feet away) with no closer or more distant
cues (for example against a clear blue sky), does it become extremely difficult
to estimate the distance? And if a reference is added at "infinity" (for
instance the moon or a cloud), how much does that help in estimating the
distance?

It sounds like there may be some value in experiments that try to
eliminate 2D depth cues, and that selectively add or remove objects
at different distances from the test object. Possibly very small flashlight
bulbs or shielded optical fibers pointed at the test subject could serve
as the test objects, in a dark environment.

John R


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

End of PHOTO-3D Digest 2124
***************************
***************************
 Trouble? Send e-mail to 
 wier@xxxxxxxxxxxxxxx 
 To unsubscribe select one of the following,
 place it in the BODY of a message and send it to:
 listserv@xxxxxxxxxxxxxxx 
   unsubscribe photo-3d
   unsubscribe sell-3d
   unsubscribe mc68hc11
   unsubscribe overland-trails
   unsubscribe icom
 ***************************