Re: stereovision question

[P3D Jim Crowell <crowell@xxxxxxxxxxxxxxxxx>]

At 10:02 PM 4/6/96, P3D John Bercovitz wrote:

>I'm
>curious how stereo vision is done in the brain.  I was just
>thinking that folks whose eyes don't track together don't have
>stereo vision.  Sounds reasonable on the surface, but why?  It
>can't be because they can't compare views at all since they still
>have two views, so it must be that somehow the retinas are
>hardwired to the same loci in the brain?  I think I've heard that
>before.  Assuming that's true, how are the comparisons between
>images made?  I would think it would start on a "pixel" level.

John,

here's something I yanked from the archive--someone's question & my reply.


>Others here have reported higher rates of stereoblindness, so I don't
>know. The other interesting part of the tape involved the development
>in infants of stereo vision. Julesz said that there is a small window
>(about 5 or 6 weeks long) somewhere around 6 months of age (I can't
>remember the details) in which an infant's brain learns stereo vision.
>If the infant doesn't have this experience (such as if it has severe
>cross-eyedness during this time period), he/she will always be stereo-
>blind for the rest of their lives. He had a device made for cross-eyed
>babies that optically provides the eyes with correct views for use >in
this training period. I have never heard of this anywhere else. >
>Ron Doerfler
>

Yes, there's a lot of relatively recent 'way cool research on how the
brain gets wired up. In adults, stereo is mediated by neurons in visual
 cortex that take inputs from specific locations (with a specific
disparity) in the two eyes. In babies, these neurons take inputs from a
range of locations in both eyes; there is a sensitive period in which
they are tuned by correlated patterns of inputs in the two eyes. This
requires that the input regions for a given cell in the two eyes point
to a common region of space. If the baby's cross-eyed enough, this
doesn't happen, & the neurons end up taking input from one eye only.
The connections from the other eye are lost, so correcting the
cross-eyedness later doesn't help.

>After local differences are noted, how is a global "picture"
>constructed?

This is not very well understood.  Certainly not at the brain
level;  there are abstract algorithm-level theories, but they're
hard to test given that we don't completely understand the early
stages...

>I know there is a lot of preprocessing done at the
>retinal level and I wonder what the picture looks like when it
>first comes to the brain on the optic nerves.

That's an enormous question, I'd have to refer you to a
textbook.  A good, recent, & relatively inexpensive one is by
Brian Wandell of Stanford;  I think it's called "Foundations of
Vision".  Highly recommended.  Unfortunately, it concentrates
pretty much on what's known about the early stages of visual
processing & I don't think it has anything on stereo.  I'll take
a look tonight, my copy's at home.

-Jim C.



------------------------------------------------
Jim Crowell
School of Optometry
360 Minor Hall
University of California
Berkeley, CA 94720-2020
(510) 642-7679
jim@xxxxxxxxxxxxxxxxx
http://john.berkeley.edu/IndividualPages/jim.html



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