The Gordonian knot

[T3D john bercovitz <bercov@xxxxxxxxxxx>]

Michael Gordon wrote up an interesting scenario in the last issue
of Stereoscopy.  He described taking a picture of a flag pole from 
near the base of it by tilting the camera up to get all of the 
pole in the view.  Then he compared the results of taking a second 
picture by use of a slide bar and by rotating the camera about the 
axis of the pole.  Very interesting scenario.
 
I'd like to extend the scenario but I'm having trouble.  Suppose 
we use a view camera to shoot the pole.  We use rise to get the 
whole pole into the scene.  Camera back is parallel to the pole.  
We us a slide bar to get the second shot.  We view the projected 
image from the correct perspective point, near the bottom of the 
screen.  The pole images are parallel but we see parallax because 
the upper ends of the pole are farther from us than the lower ends 
of the pole.  (I say "ends" because there are two images.)  So 
just as in real life, we see that the upper end of the pole is 
more distant than the lower end even though the poles are 
parallel.  Now how do we separate the pole from the stars behind 
it?  (I forgot to tell you this picture was taken at night.)  In 
particular, isn't the parallax of the upper end of the pole 
relative to the stars the same as the parallax of the lower end of 
the pole relative to the stars?  What am I missing here?
 
Thanks,
John B


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