Our method also succeeds if the number of components in
the model is orders of magnitude different from the number
of joints in the input animation, and it is robust to differences
in skeletal dimensions. It inherently handles certain morphological
differences between input skeleton and target, such
as different numbers of limbs. The approach proposes plausible
skeleton locations, despite the fact that joints are typically
not modeled explicitly in most multi-component models.
Joint correspondences between input and target skeleton
are computed in such a way that the structure of the
input skeleton is reflected, and thus, the output animation
preserves the characteristics of the input. Furthermore, our
algorithm is robust to varying input mesh resolution, and reliably
finds rigs even for input shapes made of thousands
of components. The method is generic and succeeds without
user intervention and does not requires any semantic labelling,
like in [HRE∗
08], or point-by-point correspondence,
as in [ZXTD10].