Re: Dual Video Camera Convergence Points

[Andrew Woods <RWOODSAJ@xxxxxxxxxxxxxxxx>]

Last week Ray Hannisian (ray@xxxxxxxxxxxxxx) asked about the
convergence of a pair of video cameras:

Quite a few people have commented on this but I have a few comments
which specifically relate to video.

Firstly I should point out that a lot of what I will say is covered in
an article I wrote for the SPIE Stereoscopic Displays and Applications
conference in 1993.  It is titled "Image Distortions in Stereoscopic
Video Systems" and is available in postscript format from the 
photo-3d ftp site 
ftp://bobcat.etsu.edu/pub/photo/photo-3d/technical/woods93.ps.Z
(I think that's it...?)
or alternatively in html format from my web page:
http://info.curtin.edu.au/~iwoodsa

As with most cameras, convergence of some form is necessary.  With video
cameras this can generally only be achieved in two ways:
(1) rotate the cameras inwards (toe-in) but as people have already 
    mentioned this causes keystone distortion.
(2) mount the cameras in parallel BUT
    move the camera imaging sensor (CCD for newer cameras) horizontally
    relative to the lens.  (see the paper for more detailed description
    and images)

This creates a convergence point a distance C away from the cameras.
Objects at this distance away from the cameras will appear on the
surface of the monitor.  Objects closer will appear in front
of the monitor and objects further than C will appear inside the monitor.

If the cameras were mounted parallel without the CCD shift, 'C' would
equal infinity and therefore all images would appear in front of the TV 
screen.  This is not optimal for a number of reasons.

With film cameras there are several other options for achieving the 
convergence (move the film chips when mounting, move the projectors)
but this is not possible with video without some special digital video 
processing hardware (even I don't have that - yet), so doing the
convergence at (1) or (2) is easier/cheaper, and (2) is preferable over
(1) but the amount of distortion caused by (1) may be acceptable....

As I said before, much more detailed information is available from the
paper plus lots of images illustrating the distortions.

-----------------------------------------------------------------------------
  Andrew Woods   | Centre for Marine Science and Technology,
                 | Curtin University of Technology,
      _--_|\     | GPO Box U1987, Perth W.A. 6001, AUSTRALIA.
     /      \    | Internet: WOODS_AJ@xxxxxxxxxxxxxxxx
  -->\_.--._/    | Phone:    +61 9 351 7920
Perth      v     | Fax:      +61 9 351 2377
-----------------------------------------------------------------------------


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