A 3D video scanner for face performance capture

This paper presents a 3D video scanner design for human face performance capture by active illumination stereo vision, combining a computational stereo algorithm with pattern projection to add texture to the scene. The scanner uses monochromatic video cameras to capture stereo images - a solution to obtain colour information was found by using a third web-camera and knowledge of the camera geometry to synthesise colour texture maps. Experimental results successfully demonstrate a proof of concept for an entirely computer vision based approach. The generated 3D video data being rendered and displayed using a standard graphics API yields highly realistic 3D face animation.