CagelK: Dual-Laplacian Cage-Based Inverse Kinematics

摘要

Cage-based deformation techniques are widely used to control the deformation of an enclosed fine-detail mesh. Achieving deformation based on vertex constraints has been extensively studied for the case of pure meshes, but few works specifically examine how such vertex constraints can be used to efficiently deform the template and estimate the corresponding cage pose. In this paper, we show that this can be achieved very efficiently with two contributions: (1) we provide a linear estimation framework for cage vertex coordinates; (2) the regularization of the deformation is expressed on the cage vertices rather than the enclosed mesh, yielding a computationally efficient solution which fully benefits from cage-based parameterizations. We demonstrate the practical use of this scheme for two applications: animation edition from sparse screen-space user-specified constraints, and automatic cage extraction from a sequence of meshes, for animation re-edition.