An example-based anthropometric model for face shape synthesis

This paper presents a novel method for generating varied, realistic geometric face models by synthesizing facial features according to anthropometric parameters. Our method takes as examples scanned 3D face models in order to exploit the parameter-to-geometry correlations that are presented in the real faces. A two-step model fitting approach is used to establish correspondences among scanned models. We transform the data sets of facial feature shape into vector space representations by applying a principal component analysis (PCA). For each facial feature, we compute a set of anthropometric measurements to parameterize the exemplar meshes into a measurement space. Using PCA coefficients as a compact shape representation, we approach the shape synthesis problem by forming scattered data interpolation functions that are devoted to the generation of desired shape by taking the anthropometric parameters as input.