Skip to content

← ArchivePaper2005

Skinning Mesh Animations

Doug L. James, Christopher D. Twigg

SIGGRAPHAcademic446 cites29 descendantsSkinning

Method to extract linear blend skinning weights and skeleton from animated mesh sequences, enabling compact representation of captured deformations.

Abstract

This paper extends character skinning techniques to the general setting of skinning arbitrary deformable mesh animations without requiring user-defined skeletons or bones. The method automatically estimates proxy bone transformations by applying nonparametric mean shift clustering to high-dimensional triangle rotation sequences computed via polar decomposition, then uses robust least squares to determine bone-vertex influence sets and vertex weights, preferring nonnegative least squares to avoid over-fitting. A low-rank displacement correction model defined in the rest pose provides progressive convergence with a fixed number of bones. The resulting skinned mesh animations enable hardware-accelerated rendering with matrix palette skinning, rest pose editing, level-of-detail generation, and reduced-coordinate deformable collision detection.

How to read this

Category
Method: extracting a skinned rig from mesh animation
Contributions
  • Skins arbitrary deformable mesh animations without user-defined skeletons or bones
  • Estimates proxy bone transformations via nonparametric mean shift clustering of per-triangle rotation sequences (from polar decomposition), with robust nonnegative least squares for bone-vertex influence and weights
  • Adds a rest-pose low-rank displacement correction for progressive accuracy, enabling hardware matrix-palette rendering, rest-pose editing, level of detail, and reduced-coordinate collision
Context
Generalizes character skinning (linear blend skinning / matrix-palette skinning) to a fitting problem over captured or simulated mesh animations, producing a compact skinned approximation.
Correctness
Approximation quality depends on how well the motion is captured by proxy-bone rigid transforms plus a low-rank correction; highly non-articulated or extreme deformations may need more bones or larger corrections, and the result is a fit rather than an exact reproduction.
Clarity
Accessible pipeline with well-known building blocks (polar decomposition, clustering, least squares); a first pass conveys the method, a second pass clarifies the clustering and weight-solve specifics.
How to read it
First pass for the extract-a-rig-from-animation idea and its applications; second pass on the clustering and least-squares stages if you implement skinning fitting or compression.

Builds on

Nothing in the archive, this is a starting point.

Built upon by

Related work

Keywords

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →