I'm really interested in 3D techniques which work for people without the brain mechanisms to fuse the images of their eyes (unsurprisingly).
The techniques I'm using are:
== Stereopsis: in this case, simple linear transformation in X-axis proportional to Z-axis as an approximation to a real 3D transformation
== Occlusion of one object by another: basic Z sorting, i.e. "painters algorithm"
== Subtended visual angle of an object of known size: I scale the object size proportional to its distance in the Z-axis
== Haze, desaturation, and a shift to bluishness: I interpolate the object color proportional to its distance in the Z-axis
If you can get the effect to work well, it goes from being a fairly incomprehensible jumble of grey circles (too information-dense for 2D display) to a box of bouncing balls where you can easily track individual objects as they move around.
EDIT: looking at the code, it seems I forgot about a technique that I used! I also threw in the "Vertical position" bullet point from the Wikipedia article, but without a frame of reference in the rendered scene you wouldn't be able to tell.
Edit, kind of related, here's a "virtual" stereo where I only used shadows to create the effect: http://www.flickr.com/photos/miura-ori/294022551 It works best if you use the large version of the photo.
http://picasaweb.google.com/lh/photo/MOz-Qmv8VS1pcz0zZ7D0lTo...
http://picasaweb.google.com/lh/photo/ybf2b3eh3j05J2urDrz3Lzo...
Unfortunately, the rendering speed just isn't there. The bottleneck actually seems to be the timer for running the animation, because reducing the number of objects doesn't seem to have a significant effect on the performance. I'll probably end up reworking this as an OpenGL application.
I vaguely remember a company that sold a plastic parallax barrier you could put onto touchscreen phones to do glasses-free stereoscopy. I'll post it if I find it.
Edit: http://www.spatialview.com/en/node/489 here it is, supposedly iPhone only, but I'm sure it can be hacked.