Rapid and Automatic 3D Face Modeling Using Active Appearance Models

A 3D face modeling approach for a single facial image with any pose is proposed in this paper．This paper aims to overcome the shortcomings of traditional methods such as weak universality,lack of real time,non-compatible with video transmission systems,and so on．Firstly, an improved active appearance model algorithm based on the initial position adjusting and model instance selection is applied to extract facial feature points,then we adjust candied-3wire-frame model to match global location and shape using structural characteristics of actual face and spatial affine transformation．Finally,we tune the local coordinates according to the symmetry of region of interest and construct realistic texture．Experiments results show that the proposed approach is not sensitive to focal length and the average time consumption of single image is about 300ms．