Re: Nimslo Prints vs Slides

[bercov@xxxxxxxxxx (John Bercovitz)]

>> I think to the first order, perception must follow geometry.
>> I see the aspect of perception you're probably talking about
>> as being that which fixes up the problems which come from
>> ambiguous, incomplete, and/or contradictory geometry.
 
> Reasonable, but this is a situation in which there are
> conflicting cues about the distances to the objects being
> viewed, & since the interpretation of binocular disparity
> logically depends on viewing distance, it follows that the
> stereo information for shape is ambiguous.
 
Here's another, possibly better, way of thinking about it or 
saying it.  The world has a geometry.  If distorted geometry is 
presented to the subject then to the first order the subject will 
see a distorted world.  However, the world, regardless of how it 
is presented, is seen through many powerful algorithms which can 
correct these distortions at least partially*.
 
>> I guess that's perception at work.  Most people see enlarged
>> objects if the pair is shot with a reduced baseline.  This is
>> the basis of the sport called "hypostereo".  Most people see
>> stretched objects if the viewpoint is too far from the prints.
 
> What's your basis for saying "most people"?  I find that most
> people find 3DMagic prints to look flat, but then of course I
> haven't controlled for disparities going sub-threshold...
 
Well, I've overstepped myself here.  I've never even seen 3DMagic 
prints.  What I should be saying is that pairs taken with reduced 
baseline are not all that uncommon in the specialty literature and 
that those pairs, to the first order, are perceived as having 
oversized objects in them.  Conversely, pairs taken with an 
oversized baseline are also not uncommon and objects in them 
appear miniaturized (this is hyperstereo).  In addition, if you 
free view a pair, you can see stretch/squash happen as you move 
your head farther from and closer to the pair.  These factors 
should also apply to 3DMagic prints but I don't really know enough 
about the details of that setup to say what I said earlier.
 
*I saw a most striking counter-example recently when Andrew Woods 
showed me stereo video taken with toed-in (converged) cameras.  
The hyberbolic distortion of space due to toe in (see one of 
Andrew's papers in bobcat for geometric proof and description of 
the shape of the distortion) is undeniable.  I think the reason 
for this is that as the camera pans a scene, the distortion moves 
with it and so an object becomes reshaped as it is moved from one 
part of the cameras' field to another.  This "morphing" of the 
objects in the scene is undeniable because it seems to be 
dynamically uncorrectable by the algorithms in our heads, at least 
at the rate it's presented.  If you ever had any doubt that stereo 
cameras should not have converged lenses, this videotape will make 
a believer out of you.
 
John B


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