Blending Face Details: Synthesizing a Face Using Multiscale Face Models

In this article, we present a novel approach to synthesize a new 3D face model using weighted blending of multiscale details across different faces including human (see Figure 1a) and nonhuman characters (see Figures 1b and 1c). Our approach decomposes face models into component scales, with a correspondence of salient facial features across faces. Specifically, we create a MFM that hierarchically represents the face's spatial details. A 3D face mesh is parameterized into 2D parameter space and decomposed into a base surface and multiscale continuous displacement maps (CDMs). Each MFM represents face details from coarse to fine scales while providing full correspondences across CDMs.