P3D Re: "impossible" images in stereo

["Greg Wageman" <gjw@xxxxxxxxxx>]

From: John W Roberts <roberts@xxxxxxxxxxxxxxxxx>


>The 2D "impossible images" take advantages of the fact that there's
only
>one view, in order to fool the human perceptual system in an
interesting
>way. So - can anyone think of stereo (2-view) images that take
advantage
>of the fact that there are only two views, to fool the human perceptual
>system in interesting ways?


I'm not sure I agree with your analysis, John.  Every one of these that
I can think of and that has been mentioned (the impossible fork, etc.)
uses the same 'trick'.  The image is locally consistent, but not
globally consistent.  That is, if you cover (for example) the tines of
the impossible fork, the rest of the image is 'correct', and similarly
if you cover the handle, the tines are 'correct'.  It is only the way in
which the two halves are connected that creates the ambiguity.  The
ambiguity remains because the disambiguating information has
conveniently been left out of the picture by the artist.  (E.g. the
cylindrical tine from the foreground smoothly merges into a cubic shape
in a line drawing, but in reality the laws of illumination would shade a
cylinder far differently than a cube.)

The most simple case I can think of, to answer your question, is also, I
think, the least interesting artistically (though perhaps not
intellectually).  Simply swap the left and right images to produce a
pseudoscopic image.  You now have "3D" depth cues which contradict the
2D depth cues of occlusion, perspective, etc.  Either single image is
self-consistent, but when viewed stereoscopically, appears "inside-out".
Personally I've not seen a psuedo'd pair that made me jump up and say,
"Wow!  Better than Escher!", though. :-)

That said, with the proper choice of objects and their arrangement, a
clever person might be able to come up with a pseudoscopic  pair which
is as pleasing and perplexing as one of Escher's impossible figures.
One would have to "play games" to remove disambiguating clues like
occlusion.
Lighting would also be critical.  Off the top of my head, a set of
reflective spheres of different sizes, none of which occlude, arranged
at varying distances from the camera.  In 2D, given only relative size
as the sole depth cue, the perception would probably produce a very
different idea of the scene than in 3D.  Pseudoscopically, the spheres
would then appear as concave hemispheres.  If the relative sizes and
positions were chosen carefully so they matched the correct perspective
for a set of identically-sized spheres, when combined with the
pseudoscopic depth cues....  I'm not sure what you'd have. :-)

Sounds like an interesting project for a table-top.  Or, better still, a
computer-generated image.

     -Greg W. (gjw@xxxxxxxxxx)