Given a deforming mesh in an animation, we propose a new method to segment this mesh into several near-rigid sub-meshes. From this deforming mesh over all frames of an animation, we can analyze the degree of deformation between two nearby faces on the mesh. Then, our algorithm partitions the given deforming mesh into near-rigid components where the segmentation boundaries always pass at regions of large deformation. As a result, the mesh segmentation is invariant to all frames of the given animation and the motion of faces in each near-rigid-component can be represented by the same approximate affine transformation. To demonstrate the usefulness of the algorithm, we solve the restriction of deformation transfer for triangle meshes  which requires similar reference poses between source mesh and target mesh.
All Science Journal Classification (ASJC) codes
- Computer Vision and Pattern Recognition
- Computer Graphics and Computer-Aided Design