3-DVG comments (re-post)

[T3D Jim Crowell <crowell@xxxxxxxxxxxxxxx>]

I also looked at Ken Dunkley's 3-DVG (the Initiation model).  Comments:

There are two effects: an enhancement of contrast/sharpness (probably an
effect of the reduced pupil size; if I remember right, visual acuity is an
inverted U-shaped function of pupil size, with a peak at around 2mm
diameter) and a 3-D effect.  Regarding the strength of (my) 3-D effect, as
someone else noted it's considerably weaker than a stereo pair but can be
considerably better than normal viewing of a picture.  As I suspected, the
effect appears to have two basic components:

(1) chromostereopsis: as the artificial pupils are offset from the primary
axis of the eye, different wavelengths are refracted by different amounts,
resulting in a hue-dependent stereo effect.  Two important aspects: (a)
warmer colors are shifted nearer, cooler farther, so the effect is most
pleasing when nearer objects in the scene are warm-colored and distant ones
are cooler;  (b) the disparities introduced are small.  The perceptual
effect is therefore best when looking at relatively distant scenes, when
you would expect the disparities to be small.

(2) Normally, when we view a picture there are a number of pieces of
information that tell us we're looking at something flat as opposed to a
real scene; these are sometimes referred to as 'flatness cues'.  For a
brief but clear discussion see Bruce Goldstein's _Sensation and
Perception_.  These cues include (not necessarily in order of importance):

(a) accomodation (the distance at which the lens of the eye is focused).
The resting focus of the eye is fairly distant, and it requires effort to
focus in close (as when viewing a picture in a magazine).  This effort (or
the muscular sensation produced) is detected and used by the visual system
as a crude distance cue.

=>  The 3-DVG probably reduces this cue by having the person look through
small holes.  Depth of field is increased, so the need for accomodation may
be reduced (particularly with smaller holes), and the picture may appear
farther away.  The same effect is achieved in a different way by the lenses
of a stereoscope (rather than increasing the depth of field, the distance
of best focus is shifted farther away, reducing the amount of accomodation
required).  Older people (roughly mid-forties and beyond) lose the ability
to accomodate, so they may place less reliance on the presence or absence
of this cue (which might explain why they see less of an effect with the
3-DVG).

(b) binocular disparity (stereopsis): a flat surface gives rise to a
particular pattern of horizontal and vertical disparities.  We all know
about this one.

==> When you use the 3-DVG, you only partially overlap the fields of view
from the two eyes.  In the non-overlapping portions the eyes are seeing
different things, so of course there are no binocular disparities.  In the
overlapping portions, there may be chromostereopsis (depending on the
particular picture) giving rise to disparities qualitatively appropriate
for the scene.  As noted above, these disparities are small & hence most
appropriate for distance scenes.

(c) motion parallax: as your head moves (even slightly), you get
differences in displacement of the retinal images of points at different
distances; this shearing motion is referred to as motion parallax.  When
you're looking at a picture, you don't get any motion parallax, either
between the frame and the picture or between different elements of the
picture.  This tells you it's flat.

=> The fancier 3-DVG models (which I haven't seen) include what Ken refers
to as an 'illocal frame'--essentially, each eye is looking down a tube.  If
the edges of the tube hide the edges of the picture, then you'll get motion
parallax between the new frame (the tube) and the picture, breaking the
flatness cue.  I tried improvising 'illocal frames' using two toilet-paper
roll tubes & noticed some improvement when the edge of the picture was
hidden.  If the depicted scene is distant, then you wouldn't expect
noticeable parallax between picture elements from small head motions.

It would be interesting to compare the 3-DVG effect to that of a device
invented back in the late 1800s for the same purpose (to make ordinary
pictures appear 3-D).  I can't remember what it was called, but it
basically consisted of (1) lenses to reduce accomodation, (2) a prism
arrangement that made the two eyes' viewpoints identical (eliminating
stereo), and (3) a tube at the objective end providing the 'illocal' frame.

-Jim C.


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