Popular Science reports that a Ph.D student at The University of Surrey named Zdenek Kalal has released a very cool video demonstrating an object tracking algorithm he has developed, which he calls Predator, for video image streams. You designate the object within a scene that you want to track and the algorithm can not only follow it as it moves, it also learns how the object's appearance changes as its angle or distance to the camera changes. 

Kalal outlines several possible security and machine-human interface applications for his algorithm, and it would also seem a potential fit for several camera functions, especially autofocusing.