Expression modeling for expression-invariant face recognition

摘要

Morphable face models have proven to be an effective tool for 3D face modeling and face recognition, but the extension to 3D face scans with expressions is still a challenge. The two main difficulties are (1) how to build a new morphable face model that deals with expressions, and (2) how to fit this morphable face model automatically to new 3D face scans with unknown expressions. This work presents a multi-resolution approach to semi-automatically build seven morphable expression models, and one morphable identity model from scratch. We propose an algorithm that automatically selects the proper pose, identity, and expression such that the final model instance accurately fits the 3D face scan. To prove high fitting accuracy and its use for face recognition, we perform experiments on the publicly available UND, GAVAB, BU-3DFE, FRGC v.2 datasets. Our results show high recognition rates of respectively 99%, 98%, 100%, and 97% after the automatic removal of the expressions.