Virtual Character Definition and Animation within the MPEG-4 Standard - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 32

manifold polygonal meshes of arbitrary topological type, as well as the vertex
locations. In order to support multi-resolution and fast interaction via progressive
transmission, the TS method has been combined with the Progressive Forest Split
scheme (PFS) described in Taubin (1998b). TS is used to encode the base mesh,
while the PFS being applied to encode a sequence of the forest split refinement
operations is allowed to generate higher resolutions of the base mesh. TS and
PFS approaches have been promoted to the MPEG-4 and are known as 3D Mesh
Coding (3DMC).
The second category is represented by W AVELET S UBDIVISION S URFACES (WSS),
a method recently introduced in MPEG-4 (ISOIEC, 2003). A base mesh is used
as the seed for a recursive subdivision process, during which the 3D details (i.e.,
the wavelet coefficients) needed to obtain finer and finer approximations to the
original shape are added to the new vertex positions predicted by the subdivision
scheme. WSS does not attempt to encode the base mesh—a method like TS can
be used for that purpose. Instead, the focus is on parameterization of the
underlying surface's shape over a triangular or quadrilateral base domain, in
order to obtain a multi-resolution mesh. Therefore, the main problem of these
approaches lies in finding an optimal base mesh, which is in general a computing
intensive process.
Another way of representing shapes in MPEG-4 is the M ESH G RID compression
tool (ISOIEC, 2003), which is an arbitrary, cutting plane-based representation
scheme suited for encoding the surfaces obtained from discrete 3D data sets
(e.g., 3D medical images, processed range scanner data, quadrilateral meshes,
or generic models defined by means of implicit surfaces). The resulting hierar-
chical surface representation defines the wireframe of the object's skin by: (1)
describing the connectivity between the vertices in an efficient implicit way,
called the connectivity-wireframe (CW); and (2) positioning these vertices in
relation to a regular 3D grid of points that characterizes the space inside and
outside the skin, called the reference-grid (RG). Therefore, the M ESH G RID
surface representation lies somewhat in between the two categories: (1) it has
features common to the first category in the way the connectivity-wireframe is
encoded; and (2) it exploits wavelet-based, multi-resolution techniques for
refining the shape, through the RG.
Based on these classes of geometry representation, MPEG-4 has defined
dedicated tools for the definition and the animation of virtual characters. The
next sections describe these tools in more detail.