ICA 2007

ICA 2007
7th International Conference on
Independent Component Analysis
and Signal Separation

London, UK        9 - 12 September 2007

Banner showing images of London
- Home
- Committee
- Call for Papers
- Submission
- Info for Presenters
- Dates
- Programme
- Tutorials
- Keynotes
- Papers
- Registration
- Accommodation
- Venue
- Maps
- Arrival
- Travel Tips
- Links
- Contact

Paper No: 141

Multilinear Independent Component Analysis and Dimensionality Reduction

Author(s): M. Alex O. Vasilescu, Demetri Terzopoulos

Abstract

Natural images are the composite consequence of multiple factors related to scene structure, illumination, and imaging. Independent Components Analysis (ICA) maximizes the statistical independence of the representational components of a training image ensemble, but it cannot distinguish between the different factors, or modes, inherent to image formation. We introduce a nonlinear, multifactor model that generalizes ICA. Our {\rm Multilinear ICA} (MICA) model of image ensembles learns the statistically independent components of multiple factors. We present a dimensionality reduction algorithm that enables subspace analysis within the multilinear ICA framework. As a concrete example, we consider the multilinear independent components analysis of ensembles of facial images that combine several modes, including different facial geometries (people), expressions, head poses, and lighting conditions. We demonstrate that MICA captures a set the factor subspaces, that improves face recognition.

Last Updated: 14-Aug-2007   Please read our disclaimer