Virtual Character Definition and Animation within the MPEG-4 Standard - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 63

not need a separate compression tool, in contrast to WSS for which the base
mesh is coded separately, typically with low compression tools like
I NDEXED F ACE S ET . Therefore, and together with its dedicated animation capabili-
ties, M ESH G RID is preferred over WSS for virtual character animation.
Design and Animation of a Virtual Character defined in M ESH G RID
format
Design of the Virtual Character
According to the bone-based animation requirements, the humanoid model must
consist of a global seamless mesh of the entire figure for each resolution level,
which should virtually be split into anatomical parts, e.g., shoulder, elbow, wrist,
etc., such that the motion of the skeleton can drive the appropriate parts of the
mesh.
Applying the T RI S CAN method on a humanoid model defined by implicit functions
yields as a final result a seamless mesh of the entire object at each resolution
level, as shown in Figure 13d. In order to meet the virtual-split requirement for
the mesh, the reference system for the humanoid model has to be designed
accordingly. As shown in Figure 13a and c, the reference surfaces defining the
reference grid have been chosen such that they pass through the anatomical
articulations (joints) of the body. Consequently, the single mesh is virtually split
into meaningful anatomical parts, which can be driven by the hierarchical
skeleton definition from the bones-based animation script. Notice in Figure 13a
that the density of the reference grid is higher in the areas belonging to the joints,
which will generate a denser mesh for allowing smoother deformations and
modeling. The reference system is hierarchical, providing a humanoid model with
three resolution levels, as shown in Figure 13d.
Animation of a Virtual Character defined in MeshGrid format
For the hierarchical humanoid model shown in Figure 13, the reference grid is
organized in three resolution levels: the first level contains 1,638 points, the
second level 11,056 points and the third level contains 83,349 points. The lowest
resolution level is chosen as the base level for the animation, while higher mesh
resolution levels are improving the smoothness of the deformations at the joints
and of the overall shape. It is possible, as well, to simultaneously animate
different resolution levels, in case some details only appear at higher resolution
levels. An animation sequence of the humanoid model is shown in Figure 15a. In
addition to the shaded surface mesh, the reference grid attached to the right leg
is displayed, in order to illustrate the path followed and the deformation to which
the model is constrained. Each deformation of the reference grid triggers the