My Masters research project is a 6DOF input device for biomollecular engineering and CAD modeling, using Wii Remotes as motion capture cameras ... sorta. It's not really motion capture since there's a lot more geometry involved due to ambiguity between points and often seeing <3 points in common between the cameras. Here's the poster (PDF, 348k):
This work appears as "Self-Calibrating Optical Object Tracking with Wii Remotes" in Sensors, Cameras, and Systems for Industrial/Scientific Applications X, Proc. SPIE vol. 7249. Paper (PDF, 268k), Slides (PDF, 2.2MB), Video from slides (MPEG 4, 2.3MB).
This isn't a driver so much as an abstraction of the bluetooth communication layer. In the future this will be tweaked to use TCP sockets instead of Unix Domain sockets, making the resulting communications layer uniform cross-platform (unless I hear that there's something like unix domain sockets in windows). By itself, the included app is only good for blinking your Wii Remote's LEDs idly - you need to use the wiidev_client.c API to work with the connected wii remotes. The WiiDev application is CC:BY-NC-SA, wiidev_client.c/h are CC:BY.
WiiDev.zip (85kb, source, app, API)
Significantly better understanding of the IR sensor configuration and output are documented at The Wiimote Project Wiki.
The Wii Remote's optics appear to be largely distortion free. Consistency of focal length Wii Remotes is not known yet.
The camera's on an I2C bus and appears to only need a 24MHz clock to be happy. Someone should really reverse the protocol.