graphics have become more realistic over the years in mediums such as film and
video games. Now, researchers from Microsoft's Beijing lab have taken 3D modeling
a step further with a new technique.
Microsoft's new system focuses on the human face, and surpasses even the most
advanced computer graphics used in films like. With state-of-the-art 3D scanning technology and
a motion-capture system, Microsoft researchers were able to create accurate 3D
modeling of the human face down to the very last wrinkle.
Researchers first recorded 3D facial performances by an actor using a
marker-based motion capture system. Using this recorded data, they performed a
facial analysis to identify a set of face scans needed to reconstruct certain
facial features. The idea was to determine the minimal set of face scans required
in order to make the system faster without compromising efficiency.
Microsoft researchers then built "dense consistent surface
correspondences" across these face scans, and combined the minimal set of
face scans with motion capture data in the blendshape interpolation framework
to reconstruct high-fidelity 3D facial performances.
According to results, this system is capable of capturing facial
performances matching the acquisition speed of motion capture
system and the spatial resolution of static face scans. Real and synthetic data
was observed to evaluate the system's performance.
that this system could be incorporated into Microsoft's Avatar Kinect in the
future, which uses the Kinect to
identify body position and facial expression to create an avatar on the screen
that moves along with the person controlling it. If the user were to raise an
eyebrow, the avatar would as well, for example.
The Microsoft researchers will present their new system at the SIGGRAPH 2011 computer graphics conference
in Vancouver, B.C.