Automatic extraction of the face identity-subspace

被引：11

作者：

Costen, NP ^{[1
]}

Cootes, TF ^{[1
]}

Edwards, GJ ^{[1
]}

Taylor, CJ ^{[1
]}

机构：

[1] Univ Manchester, Dept Med Biophys Imaging Sci & Biomed Engn, Manchester M13 9PT, Lancs, England

来源：

IMAGE AND VISION COMPUTING | 2002年 / 20卷 / 5-6期

关键词：

face recognition; principal components analysis; appearance models; dimensionality estimation;

D O I：

10.1016/S0262-8856(02)00004-5

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Facial variation divides into a number of functional subspaces, and variation unique to the ensemble. An improved method of measuring these is presented, within the space defined by an appearance model. Initial estimates of the subspaces (lighting, pose, identity and expression) are obtained by principal components analysis on appropriate groups of faces. A recoding algorithm is applied to image codings to maximise the probability of coding across these non-orthogonal subspaces. Ensemble-specific variation is then removed by measuring the spatial predictability of the eigenvectors excluding those, which are less predictable than the ensemble. These procedures significantly enhance identity recognition for a disjoint test set. (C) 2002 Elsevier Science B.V. All rights reserved.

引用

页码：319 / 329

页数：11

共 14 条

[1] A framework for modeling appearance change in image sequences [J].

Black, MJ ;

Fleet, DJ ;

Yacoob, Y .

SIXTH INTERNATIONAL CONFERENCE ON COMPUTER VISION, 1998, :660-667

[2]

Cootes T., 1998, Proc. ECCV, V2, P484

[3]

COSTEN NP, 1996, EUR C COMP VIS, V1, P504

[4]

COSTEN NP, 1996, 2 FACE GESTURE, P4

[5] A similarity-based method for the generalization of face recognition over pose and expression [J].

Duvdevani-Bar, S ;

Edelman, S ;

Howell, AJ ;

Buxton, H .

AUTOMATIC FACE AND GESTURE RECOGNITION - THIRD IEEE INTERNATIONAL CONFERENCE PROCEEDINGS, 1998, :118-123

[6]

EDWARDS GJ, 1997, BRIT MACH VIS C, P130

[7]

EDWARDS GJ, 1996, 2 INT C AUT FAC GEST, P328

[8] Face recognition from unfamiliar views: Subspace methods and pose dependency [J].

Graham, DB ;

Allinson, NM .

AUTOMATIC FACE AND GESTURE RECOGNITION - THIRD IEEE INTERNATIONAL CONFERENCE PROCEEDINGS, 1998, :348-353

[9] THE PRINCIPAL COMPONENTS OF NATURAL IMAGES [J].

HANCOCK, PJB ;

BADDELEY, RJ ;

SMITH, LS .

NETWORK-COMPUTATION IN NEURAL SYSTEMS, 1992, 3 (01) :61-70

[10]

LANITIS A, 1994, BMVC94 - PROCEEDINGS OF THE 5TH BRITISH MACHINE VISION CONFERENCE, VOLS 1 AND 2, P65

← 1 2 →