Now that microprocessor clock rates exceed 300MHz, SIMD integer instructions are commonplace, and shared memory multiprocessing is available for under $3,000, integration of computer vision with 3D graphics is more practical than ever before. Tracking the user's head, hands, body, and gestures could eliminate encumbering sensors. Another approach uses computer vision techniques to understand 3D structure and camera parameters in multi-view, image-based scenes and re-renders them as the user directs. Yet another uses animated characters that are aware of users and other characters. What will be the most compelling integration of computer vision with 3D graphics?
University of Toronto/Intel Corporation