…except for the LENS part… :-)

Russian tilt-shift lenses for medium-format cameras - which are probably about the absolute minimum size a projector lens would need to be - bottom out at around $US900. They’re all-manual, and you probably would need something bigger, so you’re not trying to squeeze a thousand lumens through a small aperture. Projector lens tolerances would be a bit lower (though the Russian lenses aren’t of terribly high optical quality; Canon’s rather smaller manual-focus 24mm tilt/shift lens for 35mm cameras is $US1150), but tilt/shift with good optical quality just can’t be made very cheap. And the servos to drive it couldn’t just be $5 Futabas, either.

Projectors that have optical keystone correction already have the “shift” part of a tilt/shift lens, but if you want to be able to greatly change the keystoning and focal plane then you need a lot of tilt and shift, and the more of it you need, the bigger the lens becomes, with all of its elements becoming exponentially more expensive as they grow. Add zoom and… actually, I don’t think any camera manufacturer has ever actually even tried to MAKE a tilt/shift/zoom lens. $250,000 telephotos the size of a SAM launcher, yes; tilt-shift-zoom, no. Anybody who wants to do that just puts a zoom lens on some sort of view camera, rather than try to cram all the tilt and swing stuff in the lens itself.

You can’t just stick a Lensbaby on the front of your projector :-).

The only problem with this is that it’d require a fully motorised tilt/shift/zoom lens. I’m pretty sure that bulk manufacturing could get the price of such an item down below $US3000… eventually… :-)

The two-projector splice is just beyond cool. Not too long ago, I attended a presentation where the company was showing off a workstation that used two projectors to create a single wide display. One of my reservations about the system was that it would be time-consuming to align the projectors, and that they could easily come out of alignment, even though they had a common mounting surface. This technology would very easily solve the problem.

The only problem I see is the resolution issue Dan mentioned above–I’d much prefer a system that would adjust the lens of the projector rather than the scale/skew/perspective of the image. You could still use the same test pattern to generate the 3D model of the projection surface. And if this is all built into the projector, then you wouldn’t need the computer itself for anything.

On a similar note, is there a reason why you couldn’t use four infrared LEDs, put a CCD in the projector, and ask the projector to calibrate itself Wiimote-style?