I just had the most clever idea, thanks to watching part of Firefly for the nth time.
Now and then you may need to talk to someone on subspace video relay. Each of you is looking at the monitor, not the camera, so the illusion of eye contact – and thus some gestural nuance – is lost. (Or so I assume; I have not tried it.)
Solution: bracket the monitor (horizontally) with two cameras, and transmit the image in fine vertical stripes, corresponding to an array of prisms on the monitor. Each user gets a binocular view of the other, centered in the right place.
(Later: Dad tells me he almost got to see a Sharp notebook computer with such a display. He’s jazzed about the possibility of adapting Visual Python‘s stereo feature for it.)
Drawback: the cameras are farther apart than your eyes (unless the monitor is tiny), and so the image may seem to bulge grotesquely.
Alternative: process the two views into a single interpolated view; I believe the algorithm already exists, but have no idea how much processing power it needs.