Computer scientists from Stanford University and the University of Erlangen-Nuremberg in Germany say they have created a process dubbed "live facial re-enactment" that can instantaneously transfer facial expressions.
The researchers say with the new technique, a person's smile would appear seamlessly on live video of another person's face, even if the second person is not smiling.
The team used a camera that captures gestures in three dimensions, such as the Microsoft Kinect, and developed software to map each pixel on both faces and then transfer the expression. The transfer took just 30 milliseconds, which is partly due to the software running simultaneously on several computer processors.
The researchers say live facial re-enactment could be used to improve dubbing in movies, to make video in virtual reality more realistic, and to provide instantaneous translation. For example, if an English speaker was talking to a Mandarin speaker, each would appear to be speaking the other's language.
The researchers will present their work next week at SIGGRAPH Asia 2015 in Japan.
From The New York Times
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA
No entries found