TY - GEN
T1 - Spatially smooth subspace face recognition using LOG and DOG penalties
AU - Zuo, Wangmeng
AU - Liu, Lei
AU - Wang, Kuanquan
AU - Zhang, Dapeng
PY - 2009/9/10
Y1 - 2009/9/10
N2 - Subspace face recognition methods have been widely investigated in the last few decades. Since the pixels of an image are spatially correlated and facial images are generally considered to be spatially smoothing, several spatially smooth subspace methods have been proposed for face recognition. In this paper, we first survey the progress and problems in current spatially smooth subspace face recognition methods. Using the penalized subspace learning framework, we then proposed two novel penalty functions, Laplacian of Gaussian (LOG) and Derivative of Gaussian (DOG), for subspace face recognition. LOG and DOG penalties introduce a scale parameter, and thus are more flexible in controlling the degree of smoothness. Experimental results indicate that the proposed methods are effective for face recognition, and achieve higher recognition accuracy than the original subspace methods.
AB - Subspace face recognition methods have been widely investigated in the last few decades. Since the pixels of an image are spatially correlated and facial images are generally considered to be spatially smoothing, several spatially smooth subspace methods have been proposed for face recognition. In this paper, we first survey the progress and problems in current spatially smooth subspace face recognition methods. Using the penalized subspace learning framework, we then proposed two novel penalty functions, Laplacian of Gaussian (LOG) and Derivative of Gaussian (DOG), for subspace face recognition. LOG and DOG penalties introduce a scale parameter, and thus are more flexible in controlling the degree of smoothness. Experimental results indicate that the proposed methods are effective for face recognition, and achieve higher recognition accuracy than the original subspace methods.
KW - Derivative of Gaussian
KW - Face recognition
KW - Laplacian of Gaussian
KW - Regularization
KW - Subspace analysis
UR - http://www.scopus.com/inward/record.url?scp=69849115527&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-01513-7_48
DO - 10.1007/978-3-642-01513-7_48
M3 - Conference article published in proceeding or book
SN - 3642015123
SN - 9783642015120
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 439
EP - 448
BT - Advances in Neural Networks - ISNN 2009 - 6th International Symposium on Neural Networks, ISNN 2009, Proceedings
T2 - 6th International Symposium on Neural Networks, ISNN 2009
Y2 - 26 May 2009 through 29 May 2009
ER -