Scientists from the Max-Planck Institute for Informatics at Saarland University are at CeBIT 2016 to show just how easy it can be to reconstruct faces from video recordings.
It’s no big secret any more that actors in movies often have their faces computer-edited during post-production. This requires three-dimensional facial models that are created using complex measurement processes and integrated into movie scenes by hand – or at least it did. Now the Max-Planck Institute for Informatics, Saarland University and multimedia group Technicolor are at CeBIT 2016 to exhibit a simpler and much faster solution. All they need is a standard video and the computing power to create exceptionally detailed facial models dubbed "face rigs".
This simplified process is based on a new approach to reconstructing surfaces from videos and adapting them on a computer. "We use mathematical methods to evaluate the parameters needed to capture all the details of the facial model," explains Christian Theobalt, head of the Graphics, Vision and Video group at Saarbrücken’s Max-Planck Institute and professor of computer science at Saarland University. "In addition to facial geometry (the shape of the surfaces), this includes the reflection characteristics and ambient lighting." Besides lowering costs in the movie industry, this new process could also help lend a personal touch to avatars in the virtual world, online personal assistants and virtual contacts in future telepresence applications. "Our technology can help people feel more comfortable about communicating with and through avatars," adds Theobalt.
Saarland University, Max-Planck Institute for Informatics (66123 Saarbrücken, Germany), Hall 6, Stand D28 Topic: Saarland Forschungsstand (Saarland Research Stand)