Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!ucsd!sdcc6!sdcc10!cs161agc
From: cs161agc@sdcc10.ucsd.EDU (John Schultz)
Newsgroups: comp.sources.wanted
Subject: Re: Stereoscopic computer graphics
Message-ID: <80@sdcc10.ucsd.EDU>
Date: 24 Feb 89 19:00:17 GMT
References: <147@eplrx7.UUCP>
Reply-To: cs161agc@sdcc10.ucsd.edu.UUCP (John Schultz)
Organization: University of California, San Diego
Lines: 37

In article <147@eplrx7.UUCP> cristy@vax1.acs.udel.EDU (John Cristy) writes:
>  I need references, algorithms, or source code about displaying an object
>  on a stereoscopic display. 

  You might try to find "Stereoscopic and Multiplanar Computer
Graphics", ACM SIGGRAPH '88, from August 1, 1988 Atlanta, GA. (Course
#21). 
  
  The simplest accurate equations for displaying a stereoscopic
image are [orthographic projection]:

  xl = (x + seperation)*d / z + CENTERX - seperation;
  xr = (x - seperation)*d / z + CENTERX + seperation;
  y  = y;

  There are also equations to rotate the left and right eye
perspectives, but is not recommended for scenes with large
variations in depth.

  To generate an image with a raytracer: 

    For the left eye image, set up the viewpoint coordinates for a
view slightly to the left of the center of the image.  For the right
eye, move the viewpoint to the right the same amount you moved it to
the left (simple translation).  Experiment for the best results. (in
real world coordinates, you want to simulate horizontal parallax of
about 2.5 inches max on the CRT (the ave. dist. between human
eyes)).  From my experience, best results are obtained when using
translation as above and a *slight* rotation, which would mean
keeping the center of the *camera* on the same point in space when
moving the *camera* left or right (this is in fact what I do when
digitizing an image for 3D display).


  Hope this helps,

  John Schultz