T3D Visual acuity

[Rolf Henkel <henkel@xxxxxxxxxxxxxxxxxxxx>]

Just a short comment to "visual acuity" : The receptor-spacing in the
retina
does not transfer directly to the performance of human vision, since a
lot of
tasks are accomplished by cooperative neural processing in the brain.
This is
know as hyperacuitiy. An example, not directly connect to stereo-vision,

is thealignment task of two vertical bars, somewhat like this:


           |
            |

where you have to judge if the two bars are separated vertically (like
in the
"drawing" above), or are aligned.

At such a taks, our vision system is also far better then the
receptor-spacing
in the retina would suggest. This experiment works even if the two
"bars" are
just clouds of dots, and the question is wether the two centers
of the clouds are aligned wiht each other.  Note that in such a display,

there are no edges at all, which brings me to a comment on:

> Stereo acuity is maybe ten times better but this is probably
> because the retina/brain finds edge positions to a higher
> degree of accuracy because it can average an edge over a lot
> of cones.  Edge positions may be the basis of depth perception.

This is common hype, but exactly the hyperacuity of stereovision is hard
to
explain with processes based on "edge"-positions. They are notorously
hard to
estimate to subpixel-precession. Besides, you can easily have
stereovision
(of course including hyper-acuity) in stereo displays totally lacking
edge-information.
Newer theories of stereovision (well - guess it: you can find one at my
URL listed below :-) do not use the "edge"-concept any longer.

Concerning discrepancy of retinal resolution and depth-resolution of
stereoimages:
A resolution fine enough so that it corresponds to the receptor-spacing
in the
fovea (around 1 min I guess) is fine; stereo hyperacuity is achieved not
by a higher
resolution,  but by slight variations in the greylevels of the two
stereo images.

The more grey-levels, the more intermediate depth-steps can be
perceived. A rule
of thumb (from a computational theory - might not be right for human
stereo vision) is that
the number of grey-levels in your display corresponds to the maximum
number
of intermediate depth-step which can be perceived.  Psychophysical tests
suggest
that the human visual system can indeed resolve depth-steps ten to
twenty times
finer than the foveal receptor-spacing would suggest.

Rolf Henkel

Institute of Theoretical Neurophysics, University Bremen, Germany
-----------------------------------------------------------------
Email:  henkel@xxxxxxxxxxxxxxxxxxxx
URL:    http://axon.physik.uni-bremen.de/
-----------------------------------------------------------------



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

End of TECH-3D Digest 335
*************************