3D face reconstruction with global and local constraints in double spaces

Abstract

In this paper, we proposed a new 3D face reconstruction approach to reconstruct the 3D face from a single 2D image with arbitrary pose. The proposed approach enhanced the 3D morphable model by introducing a new local face shape constraint in another space. Then, the face can be constrained in double spaces: the global face shape space and the local face structure space. The advantages of the proposed approach are: 1) it explores the global and local constraints in the double spaces; 2) it reconstructs the 3D face from a single face image with arbitrary pose. The approach is validated on BJUT_3D face database, and the 3D facial expressions are generated on FacewareHouse and FaceScurb database.