Synthesis of 3D faces using region-based morphing under intuitive control

Abstract

This paper presents a new region-based method for automatically synthesizing varied, natural looking 3D human face models by morphing local facial features according to intuitive control parameters. We automatically compute a one-to-one vertex correspondence among the unregistered face scans in a large database by deforming a generic mesh to fit the specific person's face geometry in a global-to-local fashion. With the obtained correspondence, we transform the generated data sets of feature shapes into vector space representations. We parameterize the example models using the face anthropometric measurements that reflect the facial physical structure, and predefine the interpolation functions for the parameterized example models based on radial basis functions. At runtime, the interpolation functions are evaluated to efficiently generate the appropriate feature shape by taking the anthropometric parameters as input. We use a shape blending approach for generating a seamlessly deformed mesh around the feature region boundary. The correspondence among all example textures is established by parameterizing the 3D generic mesh over a 2D image domain. The new feature texture with desired attributes is synthesized by interpolating the example textures. Our 3D face synthesis method has several advantages: (1) it regulates the naturalness of synthesized faces, maintaining the quality existing in the real face examples; (2) the region-based morphing and comprehensive face shape and texture control parameters allow more diverse faces to be generated readily; and (3) the automatic runtime face synthesis is efficient in time complexity and performs fast. Copyright © 2006 John Wiley & Sons, Ltd.