Face detection using mixtures of linear subspaces

Ming Hsuan Yang, Narendra Ahuja, David Kriegman

Research output: Chapter in Book/Report/Conference proceedingConference contribution

66 Citations (Scopus)

Abstract

We present two methods using mixtures of linear sub-spaces for face detection in gray level images. One method uses a mixture of factor analyzers to concurrently perform clustering and, within each cluster, perform local dimensionality reduction. The parameters of the mixture model are estimated using an EM algorithm. A face is detected if the probability of an input sample is above a predefined threshold. The other mixture of subspaces method uses Kohonen's self-organizing map for clustering and Fisher linear discriminant to find the optimal projection for pattern classification, and a Gaussian distribution to model the class-conditioned density function of the projected samples for each class. The parameters of the class-conditioned density functions are maximum likelihood estimates and the decision rule is also based on maximum likelihood. A wide range of face images including ones in different poses, with different expressions and under different lighting conditions are used as the training set to capture the variations of human faces. Our methods have been tested on three sets of 225 images which contain 871 faces. Experimental results on the first two datasets show that our methods perform as well as the best methods in the literature, yet have fewer false detects.

Original languageEnglish
Title of host publicationProceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000
PublisherIEEE Computer Society
Pages70-76
Number of pages7
ISBN (Print)0769505805, 9780769505800
DOIs
Publication statusPublished - 2000 Jan 1
Event4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000 - Grenoble, France
Duration: 2000 Mar 282000 Mar 30

Publication series

NameProceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000

Conference

Conference4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000
CountryFrance
CityGrenoble
Period00/3/2800/3/30

Fingerprint

Face recognition
Probability density function
Maximum likelihood
Self organizing maps
Gaussian distribution
Pattern recognition
Lighting

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition

Cite this

Yang, M. H., Ahuja, N., & Kriegman, D. (2000). Face detection using mixtures of linear subspaces. In Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000 (pp. 70-76). [840614] (Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000). IEEE Computer Society. https://doi.org/10.1109/AFGR.2000.840614
Yang, Ming Hsuan ; Ahuja, Narendra ; Kriegman, David. / Face detection using mixtures of linear subspaces. Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000. IEEE Computer Society, 2000. pp. 70-76 (Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000).
@inproceedings{55de65a8e92146deb5261df27c62fb0c,
title = "Face detection using mixtures of linear subspaces",
abstract = "We present two methods using mixtures of linear sub-spaces for face detection in gray level images. One method uses a mixture of factor analyzers to concurrently perform clustering and, within each cluster, perform local dimensionality reduction. The parameters of the mixture model are estimated using an EM algorithm. A face is detected if the probability of an input sample is above a predefined threshold. The other mixture of subspaces method uses Kohonen's self-organizing map for clustering and Fisher linear discriminant to find the optimal projection for pattern classification, and a Gaussian distribution to model the class-conditioned density function of the projected samples for each class. The parameters of the class-conditioned density functions are maximum likelihood estimates and the decision rule is also based on maximum likelihood. A wide range of face images including ones in different poses, with different expressions and under different lighting conditions are used as the training set to capture the variations of human faces. Our methods have been tested on three sets of 225 images which contain 871 faces. Experimental results on the first two datasets show that our methods perform as well as the best methods in the literature, yet have fewer false detects.",
author = "Yang, {Ming Hsuan} and Narendra Ahuja and David Kriegman",
year = "2000",
month = "1",
day = "1",
doi = "10.1109/AFGR.2000.840614",
language = "English",
isbn = "0769505805",
series = "Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000",
publisher = "IEEE Computer Society",
pages = "70--76",
booktitle = "Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000",
address = "United States",

}

Yang, MH, Ahuja, N & Kriegman, D 2000, Face detection using mixtures of linear subspaces. in Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000., 840614, Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000, IEEE Computer Society, pp. 70-76, 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000, Grenoble, France, 00/3/28. https://doi.org/10.1109/AFGR.2000.840614

Face detection using mixtures of linear subspaces. / Yang, Ming Hsuan; Ahuja, Narendra; Kriegman, David.

Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000. IEEE Computer Society, 2000. p. 70-76 840614 (Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Face detection using mixtures of linear subspaces

AU - Yang, Ming Hsuan

AU - Ahuja, Narendra

AU - Kriegman, David

PY - 2000/1/1

Y1 - 2000/1/1

N2 - We present two methods using mixtures of linear sub-spaces for face detection in gray level images. One method uses a mixture of factor analyzers to concurrently perform clustering and, within each cluster, perform local dimensionality reduction. The parameters of the mixture model are estimated using an EM algorithm. A face is detected if the probability of an input sample is above a predefined threshold. The other mixture of subspaces method uses Kohonen's self-organizing map for clustering and Fisher linear discriminant to find the optimal projection for pattern classification, and a Gaussian distribution to model the class-conditioned density function of the projected samples for each class. The parameters of the class-conditioned density functions are maximum likelihood estimates and the decision rule is also based on maximum likelihood. A wide range of face images including ones in different poses, with different expressions and under different lighting conditions are used as the training set to capture the variations of human faces. Our methods have been tested on three sets of 225 images which contain 871 faces. Experimental results on the first two datasets show that our methods perform as well as the best methods in the literature, yet have fewer false detects.

AB - We present two methods using mixtures of linear sub-spaces for face detection in gray level images. One method uses a mixture of factor analyzers to concurrently perform clustering and, within each cluster, perform local dimensionality reduction. The parameters of the mixture model are estimated using an EM algorithm. A face is detected if the probability of an input sample is above a predefined threshold. The other mixture of subspaces method uses Kohonen's self-organizing map for clustering and Fisher linear discriminant to find the optimal projection for pattern classification, and a Gaussian distribution to model the class-conditioned density function of the projected samples for each class. The parameters of the class-conditioned density functions are maximum likelihood estimates and the decision rule is also based on maximum likelihood. A wide range of face images including ones in different poses, with different expressions and under different lighting conditions are used as the training set to capture the variations of human faces. Our methods have been tested on three sets of 225 images which contain 871 faces. Experimental results on the first two datasets show that our methods perform as well as the best methods in the literature, yet have fewer false detects.

UR - http://www.scopus.com/inward/record.url?scp=33745137045&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33745137045&partnerID=8YFLogxK

U2 - 10.1109/AFGR.2000.840614

DO - 10.1109/AFGR.2000.840614

M3 - Conference contribution

AN - SCOPUS:33745137045

SN - 0769505805

SN - 9780769505800

T3 - Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000

SP - 70

EP - 76

BT - Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000

PB - IEEE Computer Society

ER -

Yang MH, Ahuja N, Kriegman D. Face detection using mixtures of linear subspaces. In Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000. IEEE Computer Society. 2000. p. 70-76. 840614. (Proceedings - 4th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2000). https://doi.org/10.1109/AFGR.2000.840614