T3D Re: 30:1

[john bercovitz <bercov@xxxxxxxxxxx>]

John Goodman asks:

> If translation is not effective, what's the drawback of tilt
> or rotation?

Best answer is to look at Andrew Woods' paper on convergence.
Convergence is the same as tilt except you tilt the camera 
instead of the object.

http://info.curtin.edu.au/~iwoodsa/

If you read his paper over, you'll see that the reconstructed
space is warped if the object is close (short perspective).  If
the perspective is very long, then there is no observable warp.

However, we do still have one problem with long perspective,
such as in the electron microscope.  That's squash.  You've
noticed how when you look through a telescope at a tree-
covered hill, the trees seem squashed together in the depth
direction.  That's perspective distortion at work.  Same thing
happens with a long telephoto shot.  The opposite happens with
a short shot with a very wide angle lens.  Here you get stretch
(of the nose out of the face in a portrait, for instance).  This 
is pertinent since taking a picture with an electron microscope 
is like taking a picture through a telescope.  It will be squashed
and there is nothing you can do to fix that other than to stand
so far back from the image that you can't see anything in the
image.  Not a solution.  So you'll see plenty of depth because
there is plenty of parallax but the reconstructed space will be
squashed.

There is a way around it.  NASA's pathfinder had a camera with
convergence so they wrote a program to build each scene in the
a computer and then set up cameras in the computer to take the 
shots correctly.  Worked wonderfully.  I think the same program 
could be applied to an electron microscope pair.

John B


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

End of TECH-3D Digest 493
*************************