Automatic, Effective, and Efficient 3D Face Reconstruction from Arbitrary View Image

In this paper, we propose a fully automatic, effective and efficient
framework for 3D face reconstruction based on a single face image in
arbitrary view. First, a multi-view face alignment algorithm localizes the
face feature points, and then EM algorithm is applied to derive the optimal
3D shape and position parameters. Moreover, the unit quaternion
based pose representation is proposed for efficient 3D pose parameter optimization.
Compared with other related works, this framework has the
following advantages: 1) it is fully automatic, and only one single face
image in arbitrary view is required; 2) EM algorithm and unit quaternion
based pose representation are integrated for efficient shape and position
parameters estimation; 3) the correspondence between 2D contour points
and 3D model vertexes are dynamically determined by normal direction
constraints, which facilitates the 3D reconstruction from arbitrary view
image; 4) a weighted optimization strategy is applied for more robust
parameter estimation. The experimental results show the effectiveness
of our framework for 3D face reconstruction.