IWT SBO 2008
Project IWT60851
Forensic Biometric Authentication
08
Fall
Department of Electrical Engineering – ESAT/PSI Kasteelpark Arenberg 10
B-3001 Heverlee (Leuven)
Michael De Smet
Demonstrator of a multicomponent 3DMM using inter-component relationships
December
09
DELIVERABLE – M12T1D1M12T1D1 IWT60851/M18T3.2D1
Document reference Authors Partners Date Confidential
IWT60851/Error: Reference source not found Michael De Smet
2009/12/15 YES
D3.2.1 Demonstrator of a multi-component 3DMM using inter-component relationships Software has been developed for the following tasks:
Automatically subdividing the face into an arbitrary number of regions, given a training database of 3D face surfaces. Internally, this software uses the k-means++ algorithm [1] for clustering.
Automatically finding the optimal per-vertex blending weights for a multi-component 3DMM starting from an initial region segmentation.
Constructing a multi-component 3DMM with inter-component relations given a training set of registered 3D surfaces and the corresponding per-vertex blending weights.
[1] D. Arthur and S. Vassilvitskii, “k-means++: The advantages of careful seeding,” in SODA, pp. 1027–1035, 2007.