H. Li, D. Huang, L. Chen, Y. Wang, J.M. Morvan
IEEE Fith International Conference on Biometrics: Theory, Applications and Systems(BTAS), Washington DC U.S.A., pp. 271-277, (2012)
In this paper, to characterize and distinguish identical twins, three
popular texture descriptors: i.e. local binary patterns (LBPs), gabor
filters (GFs) and local gabor binary patterns (LGBPs) are employed to
encode the normal components (x, y and z) of the 3D facial surfaces of
identical twins respectively. A group of facial normal descriptors are
thus achieved, including Normal Local Binary Patterns descriptor
(N-LBPs), Normal Gabor Filters descriptor (N-GFs) and Normal Local Gabor
Binary Patterns descriptor (N-LGBPs). All these normal encoding based
descriptors are further fed into sparse representation classifier (SRC)
for identification. Experimental results on the 3D TEC database
demonstrate that these proposed normal encoding based descriptors are
very discriminative and efficient, achieving comparable performance to
the best of state-of-the-art algorithms.