Face Representation and Recognition with Local Curvelet Patterns

Wei ZHOU; Alireza AHRARY; Sei-ichiro KAMATA

doi:10.1587/transinf.E95.D.3078

Face Representation and Recognition with Local Curvelet Patterns

Wei ZHOU, Alireza AHRARY, Sei-ichiro KAMATA

Full Text Views

0

Cite this

Summary :

In this paper, we propose Local Curvelet Binary Patterns (LCBP) and Learned Local Curvelet Patterns (LLCP) for presenting the local features of facial images. The proposed methods are based on Curvelet transform which can overcome the weakness of traditional Gabor wavelets in higher dimensions, and better capture the curve singularities and hyperplane singularities of facial images. LCBP can be regarded as a combination of Curvelet features and LBP operator while LLCP designs several learned codebooks from patch sets, which are constructed by sampling patches from Curvelet filtered facial images. Each facial image can be encoded into multiple pattern maps and block-based histograms of these patterns are concatenated into an histogram sequence to be used as a face descriptor. During the face representation phase, one input patch is encoded by one pattern in LCBP while multi-patterns in LLCP. Finally, an effective classifier called Weighted Histogram Spatially constrained Earth Mover's Distance (WHSEMD) which utilizes the discriminative powers of different facial parts, the different patterns and the spatial information of face is proposed. Performance assessment in face recognition and gender estimation under different challenges shows that the proposed approaches are superior than traditional ones.

Publication: IEICE TRANSACTIONS on Information Vol.E95-D No.12 pp.3078-3087

Publication Date: 2012/12/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1587/transinf.E95.D.3078

Type of Manuscript: PAPER

Category: Image Recognition, Computer Vision

Cite this

Copy

Wei ZHOU, Alireza AHRARY, Sei-ichiro KAMATA, "Face Representation and Recognition with Local Curvelet Patterns" in IEICE TRANSACTIONS on Information, vol. E95-D, no. 12, pp. 3078-3087, December 2012, doi: 10.1587/transinf.E95.D.3078.
Abstract: In this paper, we propose Local Curvelet Binary Patterns (LCBP) and Learned Local Curvelet Patterns (LLCP) for presenting the local features of facial images. The proposed methods are based on Curvelet transform which can overcome the weakness of traditional Gabor wavelets in higher dimensions, and better capture the curve singularities and hyperplane singularities of facial images. LCBP can be regarded as a combination of Curvelet features and LBP operator while LLCP designs several learned codebooks from patch sets, which are constructed by sampling patches from Curvelet filtered facial images. Each facial image can be encoded into multiple pattern maps and block-based histograms of these patterns are concatenated into an histogram sequence to be used as a face descriptor. During the face representation phase, one input patch is encoded by one pattern in LCBP while multi-patterns in LLCP. Finally, an effective classifier called Weighted Histogram Spatially constrained Earth Mover's Distance (WHSEMD) which utilizes the discriminative powers of different facial parts, the different patterns and the spatial information of face is proposed. Performance assessment in face recognition and gender estimation under different challenges shows that the proposed approaches are superior than traditional ones.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E95.D.3078/_p

Copy

@ARTICLE{e95-d_12_3078,
author={Wei ZHOU, Alireza AHRARY, Sei-ichiro KAMATA, },
journal={IEICE TRANSACTIONS on Information},
title={Face Representation and Recognition with Local Curvelet Patterns},
year={2012},
volume={E95-D},
number={12},
pages={3078-3087},
abstract={In this paper, we propose Local Curvelet Binary Patterns (LCBP) and Learned Local Curvelet Patterns (LLCP) for presenting the local features of facial images. The proposed methods are based on Curvelet transform which can overcome the weakness of traditional Gabor wavelets in higher dimensions, and better capture the curve singularities and hyperplane singularities of facial images. LCBP can be regarded as a combination of Curvelet features and LBP operator while LLCP designs several learned codebooks from patch sets, which are constructed by sampling patches from Curvelet filtered facial images. Each facial image can be encoded into multiple pattern maps and block-based histograms of these patterns are concatenated into an histogram sequence to be used as a face descriptor. During the face representation phase, one input patch is encoded by one pattern in LCBP while multi-patterns in LLCP. Finally, an effective classifier called Weighted Histogram Spatially constrained Earth Mover's Distance (WHSEMD) which utilizes the discriminative powers of different facial parts, the different patterns and the spatial information of face is proposed. Performance assessment in face recognition and gender estimation under different challenges shows that the proposed approaches are superior than traditional ones.},
keywords={},
doi={10.1587/transinf.E95.D.3078},
ISSN={1745-1361},
month={December},}

Copy

TY - JOUR
TI - Face Representation and Recognition with Local Curvelet Patterns
T2 - IEICE TRANSACTIONS on Information
SP - 3078
EP - 3087
AU - Wei ZHOU
AU - Alireza AHRARY
AU - Sei-ichiro KAMATA
PY - 2012
DO - 10.1587/transinf.E95.D.3078
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E95-D
IS - 12
JA - IEICE TRANSACTIONS on Information
Y1 - December 2012
AB - In this paper, we propose Local Curvelet Binary Patterns (LCBP) and Learned Local Curvelet Patterns (LLCP) for presenting the local features of facial images. The proposed methods are based on Curvelet transform which can overcome the weakness of traditional Gabor wavelets in higher dimensions, and better capture the curve singularities and hyperplane singularities of facial images. LCBP can be regarded as a combination of Curvelet features and LBP operator while LLCP designs several learned codebooks from patch sets, which are constructed by sampling patches from Curvelet filtered facial images. Each facial image can be encoded into multiple pattern maps and block-based histograms of these patterns are concatenated into an histogram sequence to be used as a face descriptor. During the face representation phase, one input patch is encoded by one pattern in LCBP while multi-patterns in LLCP. Finally, an effective classifier called Weighted Histogram Spatially constrained Earth Mover's Distance (WHSEMD) which utilizes the discriminative powers of different facial parts, the different patterns and the spatial information of face is proposed. Performance assessment in face recognition and gender estimation under different challenges shows that the proposed approaches are superior than traditional ones.
ER -