Synthesizing realistic facial animations using energy minimization for model-based coding

Accurate localization and tracking of facial features are crucial for developing high-quality model-based coding systems. In this paper we describe the integration of simple heuristics which are e!ective in improving the results of well-known facial feature detection with robust techniques for adapting a dynamic mesh for animation. A new method of generating a self-adaptive mesh using an extended dynamic mesh (EDM) is proposed to overcome the convergence problem of the dynamic-motion-equation method (DMM). The new method consisting of two-step mesh adaptation (called coarse-to-"ne adaptation) can enhance the stability of the DMM and improve the performance of the adaptive process. The accuracy of the proposed approach is demonstrated by experiments on eye model animation. In this paper, we focus our discussion only on the detection, tracking, modeling and animation of eye movements. 2001 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.