Virtual Character Definition and Animation within the MPEG-4 Standard - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 31

described. The next section describes in detail the first avatar animation
framework, adopted in MPEG-4 in 1998, i.e., the FBA framework. Here, the
avatar body is structured as a collection of segments individually specified by
using I NDEXED F ACE S ET . The avatar face is a unique object animated by deforma-
tion controlled by standardized feature points. The section, Virtual Characters
in MPEG-4 Part 16 , introduces a generic deformation model, recently adopted
by MPEG-4 (December, 2002), called BBA. It is shown how this model is
implemented through two deformation controllers: bones and muscles. The
generality of the model allows it to directly animate the seamless object mesh or
the space around it. Moreover, hierarchical animation is possible when consid-
ering the BBA technique and specific geometry representations, such as
Subdivision Surfaces or M ESH G RID . This advanced animation is presented in the
section, Hierarchic Animation: Subdivision Surface and M ESH G RID .
MPEG-4's Geometry Tools in a Nutshell
The simplest and most straightforward representation of 3D objects, dating from
the early days of computer graphics, is the I NDEXED F ACE S ET model. It consists in
approximating the geometry as a collection of planar polygons defined with the
aid of a list of vertex coordinates. Unfortunately, I NDEXED F ACE S ET has not been
designed to deal efficiently with highly detailed and complex surfaces, consisting
of ten to hundreds of thousands of triangles, necessary to achieve realistic
rendering of objects found in daily life. Even more important than compact
storage is the possibility to scale the complexity of the surface representations
according to the capacity of the digital transmission channels or to the perfor-
mance of the graphics hardware on the target platform. Another vital issue for
the animation of objects is the support for free-form modeling or animation,
offered by the representation method.
As a response to these new demands, several compact surface encoding
techniques have been devised during the last years. A first category of
techniques tries to respect as much as possible the vertex positions and their
connectivity as defined in the initial I NDEXED F ACE S ET description. The second
category opts for an alternative surface representation method, enabling higher
compression ratios and extra features, such as support for animation. The second
approach is more complex, certainly at the encoding stage, since a surface
described with the alternative surface representation will have to be fitted within
certain error bounds to the initial mesh description.
A representative for the first category of techniques is the Topological Surgery
(TS) representation (Taubin, 1998a), which compresses the connectivity of