Alignment problems...

[P3D Dr. George A. Themelis <fj834@xxxxxxxxxxxxxxxxxxxxx>]

The recent discussion on lens' problems and camera tilt prompted some heavy
thinking from my side.  

We can distinguish three problems/situations:

1. Different size images.
2. Vertically misaligned images.
3. Horizon out of level.

Situation #1 is the most severe.  Difference in size in the y-direction
causes vertical misalignment.  If the two chips are aligned at the bottom
then the misalignment gets worse as we move to the top.  Best strategy to
minimize bad effects of misalignment is to align the image at the center. 
Both bottom and top will be off a bit but eyestrain will be minimized.

Question:  What causes this?  Is it a difference in focal length or a
difference in the distance of each lens from the film plane?  Can it be
fixed?  In prints or during slide duplication one could correct by changing
the magnification of each copy.  But can it fixed in the camera?  If, using
shims, we change the distance of the lens from the focal plane then we can
match the sizes.  Both images will not be in focus at once but at small
apertures this might not be noticeable.  Is this a valid approach? 

Situation #2 can be corrected in mounting.  It is not a big problem, just
an inconvenience.  RBT mounts are considered a good choice to deal with
this.  A good repairman can fix the camera and align the lenses with
respect to the film which is the cause of the problem.

Situation #3 does not result in any misalignment.  I mentioned that if the
horizon must be leveled, both chips can be rotated but this will result in
real misalignment.  Why is this so?  If the image is totally flat then
there is no misalignment resulting from such rotation (flat photographers
are lucky!!!)  The misalignment in a stereo pair comes from the horizontal
displacement (parallax) present because objects are in different depth
levels.  Rotation of both chips will result in a vertical component, which
is a problem.  The situation is worse the larger the depth difference
(parallax) between two points.  In conclusion, if you have a great picture
with a slightly tilted horizon that bothers you, it can be saved if it does
not have too much depth by rotating the chips to level the horizon.

One nice example of a situation that there absolutely no horizon is to
think of a stereo looking straight down on a round table with people
sitting around it (I have taken a picture like that!)  It does not matter
how I level (rotate w/respect to its center, axis is perpendicular to film
plane) the camera because there is no preferred horizontal orientation!

I am thinking of putting all this down with some nice drawings and a few
more details in an article for Inside 3-D.  This will help clear confusion 
for these common problems that people have.  Any comments, additions,
corrections, to the above will be greatly appreciated!  -- George Themelis


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