Skip to content

← ArchivePaper2010

Easy Rigging of Face by Automatic Registration and Transfer of Skinning Parameters

Ludovic Dutreve, Alexandre Meyer, Veronica Orvalho, Saida Bouakaz

ICCVGAcademic10 citesRiggingSkinningFacial

Transfers the feature points and their point-vertex weights from an already rigged reference face onto a new mesh with radial basis functions and automatic registration, so nobody marks correspondences by hand.

Abstract

Preparing a facial mesh to be animated requires a laborious manual rigging process. The rig specifies how the input animation data deforms the surface and allows artists to manipulate a character. We present a method that automatically rigs a facial mesh based on Radial Basis Functions and linear blend skinning approach. Our approach transfers the skinning parameters (feature points and their envelopes, ie. point-vertex weights), of a reference facial mesh (source) already rigged, to the chosen facial mesh (target) by computing an automatic registration between the two meshes. There is no need to manually mark the correspondence between the source and target mesh. As a result, inexperienced artists can automatically rig facial meshes and start right away animating their 3D characters, driven for instance by motion capture data.

How to read this

Category
Method paper on automatic facial rigging by skinning transfer, from the Lyon LIRIS group with Orvalho.
Contributions
  • An automatic rig transfer that moves feature points and their point-vertex weights from an already rigged reference face to a new mesh.
  • Radial basis functions plus linear blend skinning as the deformation machinery, both cheap and already in every DCC package.
  • Automatic registration between source and target, so the user never marks corresponding landmarks by hand.
  • A stated target of letting inexperienced artists drive a fresh face with motion capture data straight away.
Context
Descends from orvalho-rig-transfer-2008 and shares its goal of authoring a rig once and reusing it. The difference is what gets transferred: Orvalho moves an anatomical rig with muscles and a skeleton, while this paper moves the far simpler payload of skinning weights and feature points, which is why it lands in real pipelines more easily.Builds on: Transferring the Rig and Animations from a Character to Different Face Models
Correctness
Linear blend skinning is the honest limit here. Whatever the reference rig could not express, the target will not express either, and the familiar candy wrapper artefacts come along with the method. The registration is automatic, which means its failure mode is silent: a target face far from the source in topology or proportion will still produce weights, just bad ones. Look for how much topological variation the examples actually cover.
Clarity
Clear and practical, written like something you could implement in an afternoon. The pipeline is laid out step by step and the notation stays light.
How to read it
First pass: abstract, pipeline figure, results. Second pass: the registration step and the weight transfer, which are the two places the method can go wrong. Third pass: only if you plan to implement it, in which case the RBF setup is the part worth copying out. The free PDF on HAL is the version to read.

Builds on

Built upon by

Nothing yet.

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 →