Adaptive Refinement for Real-time Facial Animation using Nonlinear Subdivision Scheme

Y. Zhang, E. Sung, and E.C. Prakash (Singapore)


Facial Animation, Adaptive Refinement, Nonlinear Subdivision, Real-time Simulation


This paper presents a method to adaptively refine the physically-based face model in the dynamic deformation at interactive rates. Our adaptive refinement method benefits from the subdivision schemes that are used to obtain dense and smooth data representation from sparse discrete data. We present a new subdivision scheme which performs high-quality surface smoothing through a very quick local interpolation with minimum amount of geometrical infor mation required. It is based on a combination of edge split ting operations and interpolation by blending circular arcs. The local discretization rate of the face model is determined by a quality condition that indicates where the resolution is too coarse. As the face model is deformed, the space sampling is automatically adapted to concentrate computa tional load into the regions that deform the most. Using this technique, we in effect optimize the computations, maxi mizing the overall realism while guaranteeing frame rate.

Important Links:

Go Back