Microsoft Research principal researcher Zhengyou Zhang demonstrated a photo realistic 3D avatar generated from 2D, high-quality video of a person and a simple 3D mesh model that can be created with Kinect. The avatar can be animated in real time complete with facial expressions, head movements and speech. Check out the video below:
The company also demonstrated a wedge lens which picks up hand movements above a surface using cameras that are mounted below the actual flat screen. The wedges are also used to transmit light back out the other way to create auto-stereoscopic images, which allows for glasses-free 3D, and can also be combined with head tracking from Kinect to determine where users are situated in front of a panel and re-render the content being displayed accordingly.
Although it's unclear how much of the stuff being done by Microsoft's research group will actually see the light of day in commercial products, it's still interesting to see the company is committed to developing newer forms of interacting with PCs. With Kinect's SDK coming soon, we imagine a lot more creative uses for the device will arise.