weighting depth clues

[T3D john bercovitz <bercov@xxxxxxxxxxx>]

I wonder how depth clues are weighted.  For my talk at NSA
on the geometry of perspective, I showed slide pairs of a
some cars.  The pairs were taken with normal, short, and 
long lenses.  When the audience viewed the pair taken with 
the short lens, they were sitting three times too far from 
the screen for correct perspective so the cars' geometric 
reconstruction was three times too long (stretch).  
Although the cars appeared very long, they did not appear 
three times too long.  I believe this illustrates the effect 
of the eye/brain overriding the geometric reconstruction 
with prior knowledge.  The effect of using a lens which is 
too short was dramatic in mono as well as in stereo.  I 
believe this means that the "stadiametric" system was in 
use: We know the height of the front of the car and the 
height of the rear of the car so we know their relative 
distances and from that the relative length of the car.  
(The heights of parts of the car appear unaffected by the 
stretch condition which is expected from the geometric 
reconstruction.)  So what can you tell us about how depth 
clues work together, Jim?  What overrides what under what 
conditions?  Is the internal model sometimes a stronger 
clue and sometimes a weaker clue according to how well 
we know the object?

Thanks,
John B


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

End of TECH-3D Digest 180
*************************