Re: Composing for window effects

[P3D Dr. George A. Themelis <DrT-3d@xxxxxxxxxxxxxxxxxxxxxxxxxxx>]

We continue with our P. Talbot Memorial Lecture:

>I had long suspected that large-base hypers require some sort of toe-in of
>the two cameras or shots...  At what point does toe-in become proper and 
>necessary?  

In a strict sense, it is never proper if you are viewing using a "parallel" 
system (viewer or projector with lenses aimed parallel.)  A small amount 
can be tolerated and it might be necessary when you cannot manipulate the 
stereo window.  The stereo window is set at the plane of convergence of the 
lenses.  I rarely have this problem because I mount full frame pairs in 7p 
mounts, hence I have plenty of space to set the correct stereo window.

>I can see why it is needed for aerials, but why is it needed for close-
>ups?  What is the definition of "does not overlap?"  What is the maximum 
>allowable size of the strip of image that is different?

We must distinquish between "positive" and "negative" deviation, easily
understood when the images are projected and superimposed.  If you overlap
unmounted Realist chips you will see that the images overlap at 7p which
is the location of the stereo window in the Realist camera (this is built
in the camera by slightly shifting the center of the apertures with
respect to the center of the lenses.)

Objects pasts 7ft show a positive deviation to the maximum of 1.2 mm
(for objects at infinity.)  Positive deviation is OK because it leads
to the natural impression of having the objects behind the stereo 
window.  You can identify positive deviation by the rule of 3 L's:
"Left eye sees Less in the Left side" (or, the way I use it, the
right eye sees more in the left side).  

Objects closer than 7ft show negative deviation.  Unless used properly,
negative deviation can be a problem and must be masked out.

As the camera is translated (either in a hyperstereo or a close-up), the 
field of view changes.  If you converge the lenses you will maintain the 
field of view (and the stereo window at the plane of convergence) but you 
will then introduce keystone distortion. If you do not converge the lenses 
you will introduce "ghost bands", i.e. parts of the image seen from only 
one lens and not the other, that need to be masked out.  These "ghost 
bands" are bands of negative deviation.  They do not hold any information 
and need to be eliminated to create a proper stereo window and have the 
scene unfold behind the stereo window.

When mounting the hyperstereo or close-up, the purpose is to eliminate 
this negative deviation and introduce some positive deviation (to a 
maximum of 1.2 mm in Realist format mounts) for the stereo window.

All this sounds sounds complicated but it is easily understood if you
draw a few simple drawings or experiment in mounting or projection.

-- George Themelis


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