Automatic 3D facial wire-frame model adaptation.
Lu Yu · Journal of Zhejiang University(Sciences Edition) · 2004
3D facial wire-frame model is an important method for the natural facial representation. How to adapt the general model to the special human face deals with the two main routes including automatic extraction of facial feature points and adaptation of the model. In this paper, the extracted face candidates from video scene with complex background by the algorithm of skin-color segmentation and morphological filtering are adapted through their directions of the region inertial ellipses and verified by prior face template to find out real face regions and locate the responding facial feature regions. With the prior knowledge of the color properties and spatial distribution of the facial feature region, a feature-based method of face detection is employed to extract some of the essential facial feature points and the edges of the facial feature are fitting by curves and active contour model, from which the facial points are extracted and serve for the estimation of the face pose and the interpolation node of radial basis function interpolation for automatic adaptation of the 3D facial wire-frame model.