Shape-based detail-preserving exaggeration of extremely accurate 3D faces

摘要

We present an approach to automatically exaggerate the distinctive features of extremely detailed 3D faces. These representations comprise several million triangles and capture skin detail down to the pores. Despite their high level of realism, their size makes visualization difficult and realtime mesh manipulation infeasible. The premise of our methodology is to first remove the detail to obtain low resolution shape information, then perform shape-based exaggeration on a low resolution model and finally reapply the detail onto the exaggeration to recover the original resolution. We also present the results of applying this methodology to a small set of faces.