Game Development Reference
In-Depth Information
expressions. As difficult as the face is to model, it is even more problematic to
animate. Facial deformations are a product of the underlying skeletal and
muscular forms, as well as the mechanical properties of the skin and subcutane-
ous layers, which vary in thickness and composition in different parts of the face.
The mouth area is particularly demanding, because there are additional move-
ments of the mandible and intra-oral air pressures, which influence the visible
morphology of this area. All of these problems are enormously magnified by the
fact that we as humans have an uncanny ability to read expressions and lips —
an ability that is not merely a learned skill, but part of our deep-rooted instincts.
For facial expressions, the slightest deviation from reality is something any
person will immediately detect. This said, people would find it difficult to put their
finger on what exactly it is that was wrong. We have to deal with subtle effects
that leave strong impressions.
Face animation research dates back to the early 70s (Parke, 1972). Since then,
the level of sophistication has increased dramatically. For example, the human
head models used in Pixar's Toy Story had several thousand control points each
(Eben, 1997). More recent examples, such as Final Fantasy and Lord of the
Rings, demonstrate that now a level of realism can be achieved that allows
“virtual humans” to play a lead part in a feature movie. Nevertheless, there is still
much manual work involved.
For face animation, both 2D image-based and 3D model-based strategies have
been proposed. Basically, the choice was one between photorealism and
flexibility.
2D: For reaching photorealism, one of the most effective approaches has been
to reorder short video sequences (Bregler et al., 1997) or to 2D morph
between photographic images (Beier et al., 1992; Bregler et al., 1995; and
Ezzat et al., 2000). A problem with such techniques is that they do not allow
much freedom in face orientation, relighting or compositing with other 3D
objects.
3D: A 3D approach typically yields such flexibility. Here, a distinction can be
made between appearance-based and physics-based approaches. The
former is typically based on scans or multi-view reconstructions of the face
exterior. Animation takes the form of 3D morphs between several, static
expressions (Chen et al., 1995; Blanz et al., 1999; and Pighin et al., 1998)
or a more detailed replay of observed face dynamics (Guenter et al., 1998;
Lin et al., 2001). Physics-based approaches model the underlying anatomy
in detail, as a skull with layers of muscles and skin (Waters et al., 1995;
Pelachaud et al., 1996; Eben, 1997; and Kähler et al., 2002). The activation
of the virtual muscles drives the animation. Again, excellent results have
been demonstrated. Emphasis has often been on the animation of emotions.
Search WWH ::




Custom Search