Facial expression synthesis by radial basis function network and image warping

摘要

This paper presents a novel approach in synthesizing various facial expressions by image warping, using spatial displacement of facial landmark point generated by radial basis function (RBF) networks. We trained two RBF networks from test images to obtain these spatial displacement. One RBF network can be used to generate different degrees of expressions, while the other can be used to obtain mixed facial expressions. We discuss the method used and demonstrate the results.