Re: Custom 3D video cameras

[P3D Andrew Woods <andrew@xxxxxxxxxxxxxxxxxxxxx>]

Ray Hannisian  <ray3d@xxxxxxxxx> wrote: 
> My solution to the "window problem" is to 'toe-in' the cameras, creating a
> convergence point which varies according to the subject matter I am
> shooting, the interocular positioning of the cameras and whether or not a
> distant background is visable in comparison with close-up objects.  The
> distortions created by this procedure are better than having infinity
> objects appear in front of the TV screen, and having unresolvable edges.

Video cameras generally need to be converged by some method to generate
usable 3D video.  The convergence can be achieved either by the usual
"toe-in" method or by keeping the cameras parallel but also introducing image
shift either by mechanically shifting the CCD or doing an electronic
shift of the video images.  The toe-in method creates distortions 
(keystone distortion and depth curvature) whereas the parallel with image
shift method does not.  

I wrote a paper called "Image Distortions in Stereoscopic Video 
Systems" which analyses these distortions.  I've placed it on the web
"http://info.curtin.edu.au/~iwoodsa".  It has some pretty heavy math
but should answer some of your questions.

Parallel cameras without image shift would be a disaster.

> "Stereographics Corp." uses electronic circuitry in their products to
> produce a 'side-slipping' of the images to create what they call
> "coincidence", rather than "convergence".  This is equivilant to shooting
> film with parallel cameras and positioning the convergence (coincidence)
> point during mounting.

I agree that the use of the word convergence can be a bit confusing.
I seem to remember someone using a really neat term when at the last
Stereoscopic Displays and Applications conference.  I can't remember
what it was now but it may well have been coincidence...?

As you say StereoGraphics use parallel cameras with an electronic shift
of the images.  Unfortunately this results in black bands (albiet small)
on either side of the screen.  That's why I went for a mechanical CCD shift.

> I am looking for a mechanical solution.
> 
> I would like to separate the lens-boards from the cameras in such a way that
> they may be offset sideways while remaining parallel to the CCDs, as is done
> in 4x5 view cameras.  A small screw adjustment should make it possible to
> 'dial' the correct displacement.  A bellows arrangement would keep it
> 'light-tight'.  

I use a mechanical assembly in my latest underwater stereoscopic video
camera to shift the CCDs relative to the lenses.  The left CCD is shifted
to the left relative to the left lens and the right CCD is shifted right
relative to the right lens.  The lenses are kept parallel and the CCDs
are kept perpendicular to the axes of the lenses.
You can see the camera on my web site
"http://info.curtin.edu.au/~iwoodsa/cscs"
Sorry, no peeking inside.  :-(

> I would appreciate advice from anyone with some optics background on the
> following matters:
> 
> When dealing with sideways displacement, will a standard lens provide
> coverage outside of its normal target zone?  (I've always thought that view
> camera lenses were built to cover more area than a non-movable lens.)

Depends on the particular lens you have.  Every lens will be different.
Some fairly  simple experiments will determine what coverage a particular
lens has.

> Will 'side-slipping' my lenses create a different F-stop, (reducing the
> amount of light reaching the target)? 

Probably not but it depends how wide the coverage of the lens is and whether
there is any vingetting towards the edges.  (is that the term?  I know it 
isn't the spelling :-)
  
The amount of shift will be fairly small anyways so it will only
be the very edge which will be affected if there is any vingetting 
(darn, there's that term again).

> Should the lenses retain the original inter-occular distance, and the
> cameras be displaced, or vice versa?  (My intuition leans toward the lens
> positions determining the correct spacing.) 

The small amounts of movement you are talking about, it won't matter.
Do whichever is easiest.

> For the moment, this is only a hypothetical camera system.  However, I
> eventually wish to build a portable Betacam SP video system based upon two
> Toshiba IK-TU40A 3-chip miniature cameras and a SONY BVV-5 recorder.  I
> imagine it having variable inter-occular spacing, convergence and
> coincidence and synched zoom lenses.  

Good luck.  There should be lots of people interested if you succeed
(oh, and if it's cheap..  ;-)

Regards,
Andrew.

+------------------------------------------------------------------+
|  Andrew Woods   | Centre for Marine Science and Technology,      |
|                 | Curtin University of Technology,               |
|       _--_|\    | GPO Box U1987, Perth W.A. 6001, AUSTRALIA.     |
|      /      \   | Email: A.Woods@xxxxxxxxxxxxxxxxxx              |
|   -->\_.--._/   |     or Andrew@xxxxxxxxxxxxxxxxxxxxx            |
| Perth      v    | WWW: http://info.curtin.edu.au/~iwoodsa        |
|                 | Phone: +61 9 351 7920     Fax: +61 9 351 2377  |
+------------------------------------------------------------------+


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