Facial Animation - Computer Animation: Algorithms and Techniques - page 321

10.2.1 Creating a continuous surface model
Creating a model of a human head from scratch is not easy. Not only must the correct shape be gen-
erated, but when facial animation is the objective, the geometric elements (vertices, edges) must be
placed appropriately so that the motion of the surface can be controlled precisely. If the model is dense
in the number of geometric elements used, then the placement becomes less of a concern, but in rel-
atively low resolution models it can be an issue. Of course, one approach is to use an interactive system
and let the user construct the model. This is useful when the model to be constructed is a fanciful crea-
ture or a caricature or must meet some aesthetic design criteria. While this approach gives an artist the
most freedom in creating a model, it requires the most skill. There are three approaches used to con-
struct a model: refining from a low-resolution model, modifying a high-resolution simple shape, and
designing the surface out from high-resolution areas.
Subdivision surfaces (e.g., [ 9 ] ) use a polygonal control mesh that is refined, in the limit, to a smooth
surface. The refinement can be terminated at an intermediate resolution, rendered as a polygonal mesh,
vertices adjusted, and then refined again. In this manner, the designer can make changes to the general
shape of the head at relatively low resolution. Figure 10.6 shows the initial stages of a subdivision-
based facial design [ 27 ]. Subdivision surfaces have the advantage of being able to create local com-
plexity without global complexity. They provide an easy-to-use, intuitive interface for developing new
models, and provisions for discontinuity of arbitrary order can be accommodated [ 9 ]. However, they
are difficult to interpolate to a specific dataset, which makes modeling a specific face problematic.
Alternatively, the designer may start with a relatively high-resolution simple shape such as a sphere.
The designer pushes and pulls on the surface to form the general features of the face and refines areas as
necessary to form the face [ 17 ] (see Figure 10.7 ) .
Another alternative is to build the surface out from the expressive regions of the face—the mouth
and eyes. This approach can be used to ensure that these regions are flexible enough to support the
intended animation. See Figure 10.8 for an example of an initial surface design [ 29 ] .

Fig 1. Fig 2. Fig 3.

Fig 4. Fig 5. Fig 6.

Fig 7. Fig 8. Fig 9.

FIGURE 10.6
Early stages in facial modeling using subdivision surfaces [ 27 ] .