Game Development Reference
In-Depth Information
I have implemented this approach inside the commercial modeling package,
Maya. It has several advantages over other techniques I am aware of. The cost of
recomputing the spline curve is very minimal and only done once per frame (the
curve can be sampled at a constant density so that u ( t ) can be computed as the
weighted blend of neighboring points).
The matrix-blending equation in smooth skinning is replaced with the cre-
ation of a single transformation matrix centered at u ( t ) for each vertex. This is
generally faster and involves computing the position on the curve at u ( t ), its first
derivative u ( t ), and a cross-product.
The results under twisting deformations are impressive. The mesh vertices
remain at a fixed distance from the bones and will never form a candy wrapper.
Adjusting the positions of the control points of the curve (sliding them along
the length of the bones) can be done to create longer or tighter bending between
joints. Additionally, there is nothing preventing you from treating the curve as
just another joint so that contributions from the curve may be blended with other
joints in the smooth skinning algorithm.
Beyond skinning.... I think spline-skinning is an exciting and remarkably sim-
ple way of resolving the problems inherent in the smooth skinning algorithm. It
is likely that some variant of the smooth skinning process will be in use long into
the future. It is worth investing the time and resources to augment or extend the
algorithm so that it works the way artists expect it to, that is, without collapsing
into a rubber hose.
That said, this deformation method alone will never be sufficient in capturing
all of the subtleties of a creature's skin. Fortunately, we can use it as a base
upon further deformations may be added to account for things like skin stretching,
wrinkles, and collisions with underlying bones or muscles.
13.4 Anatomical Collisions
One of the greatest problems in designing creature deformations is that of main-
taining volume. We don't need to measure the exact volume and literally maintain
this over time; that would be overkill. Instead, it is sufficient to create the impres-
sion that there is flesh, muscle, and bone underneath a creature's skin.
A limb that bends like a rubber hose is very alarming to human eyes. A
limb that shows compression and bulging will immediately give the impression of
underlying mass. There are existing techniques that create an inner scaffolding of
springs that resist compression to maintain volume. Interested readers should see
[ZHO] for an excellent introduction to these interior lattice methods. The method