Inverse kinematics for reduced deformable models
ACM SIGGRAPH 2006 Papers
Smooth contact-aware facial blendshapes transfer
Proceedings of the Symposium on Digital Production
| Hi-index | 0.00 |
Computer-generated character animation, where human or anthropomorphic characters are animated to tell a story, holds tremendous potential to enrich education, human communication, perception, and entertainment. However, current animation procedures rely on a time consuming and difficult process that requires both artistic talent and technical expertise. Despite the tremendous amount of artistry, skill, and time dedicated to the animation process, there are few techniques to help with reuse. Although individual aspects of animation are well explored, there is little work that extends beyond the boundaries of any one area. As a consequence, the same procedure must be followed for each new character without the opportunity to generalize or reuse technical components. This dissertation describes techniques that ease the animation process by offering opportunities for reuse and a more intuitive animation formulation. A differential specification of arbitrary deformation provides a general representation for adapting deformation to different shapes, computing semantic correspondence between two shapes, and extrapolating natural deformation from a finite set of example shapes. Deformation transfer adds a general-purpose reuse mechanism to the animation pipeline by transferring any deformation of a source triangle mesh onto a different target mesh. The transfer system uses a correspondence algorithm to build a discrete many-to-many mapping between source and target triangles that permits transfer between meshes of different topology. Results demonstrate retargeting both kinematic poses and non-rigid deformations, as well as transfer between characters of different topological and anatomical structure. Mesh-based inverse kinematics extends the idea of traditional skeleton-based inverse kinematics to meshes by allowing the user to pose a mesh via direct manipulation. The user indicates the class of meaningful deformations by supplying examples that can be created automatically with deformation transfer, sculpted, scanned, or produced by any other means. This technique is distinguished from traditional animation methods since the expensive character setup stage is avoided. It is distinguished from existing mesh editing algorithms since the user retains the freedom to specify the class of meaningful deformations. Results demonstrate an intuitive interface for posing meshes that requires only a small amount of user effort. (Copies available exclusively from MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)