Image-based 3D model adaptation for face reconstruction and expression animation

An image-based model adaptation method was adapted for realistic facial model reconstruction and expression animation. Firstly, 93 feature points were extracted from both front and side images of an individualized face. Secondly, global geometrical adaptation was performed according to the correspondence of the feature points and the generic 3D facial model, and then, 3D generic facial model was transformed with the Radial Basis interpolation to fit the specific facial image as a local adaptation. After those, texture mapping was used to reconstruct 3D specific facial model, and texture image was generated by multi-resolution mosaic. Subsequently, the muscular model and elastic repulsion model were employed for generating facial expression animation and retaining facial topological. Dynamic local subdividing on mesh triangles was used for enhancing reality.