Coupled 3D reconstruction of sparse facial hair and skin applied to various different facial hairstyles. |
Researchers
at Disney Research, Zürich, ETH Zürich, and Cornell University have
invented a system to digitize facial hair and skin. Capturing facial
skin and geometry is a fundamental technology for a variety of
computer-based special effects for movies. Conventional face capturing
is well established and widely utilized in the entertainment industry to
capture a three-dimensional model of an actor’s face. However, up to
now, no method was capable of reconstructing facial hair or even
handling it appropriately. This omission is surprising as facial hair is
an important component of our popular culture.
The
system developed by the Disney Research Laboratory in Zürich
constitutes a significant technical breakthrough in the field of
digitizing human faces and was presented at ACM SIGGRAPH, the
International Conference on Computer Graphics and Interactive
Techniques.
“Our
method captures individual strands of facial hair and stores them
separately from the actual human face. This approach allows us to
‘shave’ people with facial hair virtually with the computer,” said Thabo
Beeler, a computer scientist at Disney Research, Zürich, who is the
main inventor of the technology.
The
system employs several consumer-grade photo cameras to capture a face
in a fraction of a second. The method then automatically detects hairs
in the captured images. These hairs are being tracked and followed in
the input images, much like we follow a path on a map with our fingers. A
mathematical method called multi-view stereo (MVS) reconstructs them in
three dimensions. The trick the researchers applied is to remove the
hair strands from the input images similar to an artist painting over
parts of a picture. This process makes the 3D skin surface to look as if
it were digitally shaved. The system was applied to a large variety of
different facial hair styles, ranging from designer stubbles all the way
to wild mustaches, to demonstrate its robustness. The produced results
look very compelling.
Prof.
Markus Gross, director of Disney Research, Zürich, stated, “The
long-term goal of our research is to make facial animation and special
effects more realistic and ultimately indistinguishable from reality.
This method is going to be a very important step toward this long term
goal”.
In
addition to Prof. Gross and Thabo Beeler, the participants on the
project were Bernd Bickel (Disney Research, Zürich), Gioacchino Noris
(Disney Research, Zürich and ETH Zürich), Paul Beardsley (DRZ), Steve
Marschner (Cornell University) and Robert W. Sumner (Disney Research,
Zürich).
Source: Disney Research
