Expression Invariant 3D Face Recognition with a Morphable Model
Author
Brian Amberg
brian.amberg@unibas.ch
Reinhard Knothe
reinhard.knothe@unibas.ch
Thomas Vetter
thomas.vetter@unibas.ch
Abstract
We describe an expression-invariant method for face recognition by fitting an identity/expression separated 3D Morphable Model to shape data. The expression model greatly improves recognition and retrieval rates in the uncooperative setting, while achieving recognition rates on par with the best recognition algorithms in the face recognition great vendor test. The fitting is performed with a robust nonrigid ICP algorithm. It is able to perform face recognition in a fully automated scenario and on noisy data. The system was evaluated on two datasets, one with a high noise level and strong expressions, and the standard UND range scan database, showing that while expression invariance increases recognition and retrieval performance for the expression dataset, it does not decrease performance on the neutral dataset. The high recognition rates are achieved even with a purely shape based method, without taking image data into account.
Conceptions:
- Registration: Registering two surfaces means finding a mapping between a template surface and a target surface that describes the position of semantically corresponding points.
- Dense registration: methods find a mapping from each point in the template onto the target while sparse methods find correspondence only for selected feature points. We present a dense registration method.
- Regularisation:
- To choose the “correct” deformation from all possible warps, a registration algorithm has to impose constraints on the deformation. In this context, this is called regularisation of the deformation field.
- for here: minimizing the difference between transformations acting on neighbouring vertices of a mesh.
Fitting
The main difference, is that the deformation model is a statistical model and the optimisation in each step is an iterative method, which finds the minimum of a convex function.
- steps:
Repeat until convergence:- Find candidate correspondences by searching for the closest compatible point for each model vertex.
- Weight the correspondences by their distance using a robust estimator.
- Fit the 3DMM to these correspondences using a regularization strength of $\theta_i$.
- Continue with the lower $\theta_{i + 1}$ if the median change in vertex position is smaller than a threshold.
Nonrigid optimal step ICP algorithms
Deformation
