Advanced Video Processing

Digital video editing and computer graphics have allowed unbridled creativity for professional film makers. However, the software tools and techniques used to accomplish these tasks require expert knowledge and are often slow. With this project, we aim to improve existing video processing tools which currently slow down existing users and block new users from the creative process.

One particularly difficult task in advanced video processing is seamlessly changing the appearance of objects, or changing the lighting in scenes. Typically this requires painstaking manual painting, which is laborious, or it requires significant compute resources, taking many hours for just a few seconds of footage. We are developing a new technique which provides fast interactive tools for video editing. The video is split into two component parts: the reflectance, or color, of scene objects, and the scene lighting. So-called 'intrinsic video decomposition' was previously very slow, but this new human-assisted approach speeds up the process by a factor of 100. This greatly quickens the creative process, and so opens the door to more sophisticated video content editing of appearance and illumination.

Team members: Steffen Kirchhoff, James Tompkin, Deqing Sun, Hanspeter Pfister