As-Rigid-As-Possible Character Deformation Using Point Handles

In this paper, we present a versatile, point handles based character skinning scheme. Point handles are easier to design and fit into an object's volume than a skeleton. Moreover, with a conventional blending technique such as linear blending, point handles have been successfully demonstrated to handle stretching, twisting, and supple deformation, which are difficult to achieve with rigid bones. In the context of only blending, limbs, however, are not bent rigidly with point handles, limiting the space of possible deformations. To address this, we propose a method that automatically recovers the local rigidities of limbs via minimizing a surface-based, nonlinear rigidity energy. The minimization problem is subjected to the positions of a set of point handles' proximal vertices. The positions fitting point transformations are computed by linear blend skinning, leading to speedups of the minimization in particular for large deformations. The use of nonlinear energy also allows versatile posing by intuitively selecting which point handles provide their proximal vertices on-the-fly. The degrees of freedom in modeling user constraints are reduced, and the skinning process is automated by relevant functionalities included in our scheme. The effectiveness of our scheme is demonstrated by a variety of experimental results, showing that the scheme could be an alternative to skeletal skinning.