A Feature-Preserved Simplification for Autonomous Facial Animation from 3D Scan Data

Abstract

We propose a new simplification algorithm of facial models for animation. For the facial animation, the models are often simplified from complex scan data based on geometric features, but it leads to decrease the quality and such features are easily noticed by human perception. For example, a lip line and eyebrows easily lose their details by geometry-based simplification. In this paper, facial features are extracted using an image processing of a 2D texture image and the curvature analysis of the 3D geometry, which improves the details around the feature areas of the facial model. Especially if lip contact line is simplified to one or two edges, it may not be proper for lip animation. Finally, we will show that our simplified model can produce as good as a facial animation as the one from the original model.