"Converging" fields of view

[P3D <Gregg_Podnar@xxxxxxxxxxxxxxxxx>]

Neil Harrington wrote:

>How is the adjustment done that "sets the stereo window"?  If you have some
>method _other_ than converging the fields of view, I'd sure like to know
>what it is.  Seriously.

If you define the "field of view" as the pyramid within which points will be
imaged on the film, then you may think of these pyramids as converging.  This
contains little information about the geometry of the cameras and is why we
avoid the use of the term.

If you toe-in cameras, then there will be at most one vertical line segment
of points which share the same geometric relationship (are coincident) in 
both images (due to keystone distortion).

If you mount cameras parallel, there are an infinite number of plane segments
for which a one-to-one correspondence can be found on the film planes.  
Placing rectangles of film (video sensor) centered behind each lens puts the 
stereo window at infinity.  As the chosen lenses cover more area than the 
film, the film chips may be placed off-center, which defines a different 
plane segment of coincident points.

Note that for a realistic view, the displaying system must maintain a 
relationship with the recording system (i.e. both chips may be shifted to
the right of each lens axis ==> the viewing screen must be shifted 
appropriately to the left of the viewer (parallel to the line between the 
eyes)).  It is easier to consider films which are infinite in size, an
infinite display screen, and aperatures at the film planes which each select a
portion of an image, presented on the appropriate portion of the screen.


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