Estimation of asymmetry parameters for mesh based facial models
Pujitha Gunaratne, Yoichi Sato · 2004
Construction of human like avatars is a key to produce realistic animation in virtual reality environments and has been a commonplace in present day applications. However most of the models proposed to date intuitively assume human face as a symmetric entity. Such assumptions produce unfavorable drawbacks in applications where the analysis and estimation of facial deformation patterns play a major role. Thus in this work we propose an approach to define asymmetry parameters of facial expressions and a method to evaluate them. The proposed method is based on capturing facial expressions in three-dimension by using a rangefinder system. Three-dimensional range data acquired by the system are analyzed by adapting a generic mesh with facial topology. The asymmetry parameters are defined based on the elements of the generic mesh and evaluated for facial expressions of normal subjects and patients with facial nerve paralysis disorders. The proposed system can be used to store asymmetric details of expressions and well fitted to remote doctor-patient environments.