Rapid modeling of 3D faces for animation using an efficient adaptation algorithm

Abstract

This paper presents a new efficient method for the reconstruction of a personalized 3D facial model for animation from range data. Our method adapts a generic control model with anatomical structure to the geometry of a specific person's face with minimum manual intervention. The face adaptation algorithm starts with the specification of a small set of anthropometric landmarks on the 2D images of both the generic control model and individual face. 3D positions of landmarks are recovered automatically by using a projection-mapping approach. A global adaptation is then carried out to adapt the size, position and orientation of the generic model in the 3D space based on a series of measurements between the recovered 3D landmarks. After the global adaptation, a local adaptation deforms the generic model to fit all of its vertices to the scan data-set. The underlying muscle structure is automatically adapted as well, such that the reconstructed model not only resembles the individual face in shape and color but also reflects the structure of human face including skin and muscles, therefore can be animated immediately with the given muscle parameters.