Maddy

I did quite a bit of research a few years ago on creating discrete left and right images for real-time red/cyan anaglyph conversion, either as a parallel pair with the left image converted to grayscale and the right one as a full color image, or a parallel pair with the left image converted to red and the right one as a cyan image… but with all the shade corrections, hue adjustments, etc, needed for a *perfect* anaglyph.
I archived every bit of the years of research, along the way, in the 3DTV group, if you are willing to dig around.
You can create these yourself with VirtualDub freeware. I have an old tutorial, here (scroll down to “Shade-Corrected, Parallel Pairs For Anaglyph”).
You can also convert these to field sequential 3D with one set of lines being one perspective and the other set of lines being the other perspective. This section on the same page tells the steps: Red-Cyan, Shade-Corrected, Color Anaglyph - Field Sequential (also called “Interleaved” or “Interlaced”)
My original purpose for doing this was to avoid anaglyph compression artifacts, which are really nasty with lossy video compression.
I’m sure it has a good use somewhere… somehow… maybe more so these days with the available digital displays and such. You can page-flip them, play them field sequentially, overlay them, etc… all with no compression artifacts. The discrete pairs can be compressed with any good lossy compression, e.g. Xvid is as good as any.
I have more recently solved the problem of anaglyph compression artifacts by animating the (screen) parallax and controlling (minimizing) the stereoscopic deviation. If you can make a YouTube anaglyph look good, then you are on the right track (some are better than others… ;-))
See:
http://www.youtube.com/RogerTubed