A subspace method for maximum likelihood target detection

Author

Moghaddam, Baback ; Pentland, Alex

Author_Institution

Media Lab., MIT, Cambridge, MA, USA

Volume

3

fYear

1995

fDate

23-26 Oct 1995

Firstpage

512

Abstract

We present an unsupervised technique for visual target modeling which is based on density estimation in high-dimensional spaces using an eigenspace decomposition. A computationally efficient and optimal estimator for a multivariate Gaussian distribution is derived. This density estimate is then used to formulate a maximum likelihood estimation framework for visual search and target detection. Our learning technique is applied to the probabilistic visual modeling and subsequent detection of facial features and is shown to be superior to matched filtering

Keywords

Gaussian distribution; Gaussian processes; eigenvalues and eigenfunctions; face recognition; image recognition; maximum likelihood detection; maximum likelihood estimation; unsupervised learning; computationally efficient estimator; density estimation; eigenspace decomposition; facial features detection; high dimensional spaces; image processing; learning technique; maximum likelihood estimation; maximum likelihood target detection; multivariate Gaussian distribution; optimal estimator; probabilistic visual modeling; subspace method; unsupervised technique; visual search; visual target modeling; Covariance matrix; Face detection; Filtering; Karhunen-Loeve transforms; Matched filters; Maximum likelihood detection; Maximum likelihood estimation; Object detection; Principal component analysis; Space technology;

fLanguage

English

Publisher

ieee

Conference_Titel

Image Processing, 1995. Proceedings., International Conference on

Conference_Location

Washington, DC

Print_ISBN

0-8186-7310-9

Type

conf

DOI

10.1109/ICIP.1995.537684

Filename

537684