PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Inlier-based ICA with an application to super-imposed images
Frank Meinecke, Stefan Harmeling and Klaus-Robert Müller
International Journal of Imaging Systems and Technology 2005.

Abstract

This paper proposes a new ICA method which is able to unmix overcomplete mixtures of images. Furthermore, the method is designed to be very robust against outliers, which is a favorable feature for ICA algorithms since most of them are extremely sensitive to outliers. Our approach does not robustify an existing algorithm by some outlier detection technique. Instead we show that a simple outlier index can be used directly to solve the ICA problem for super-Gaussian source signals. Our inlier-based ICA (abbr. IBICA) is outlier-robust by construction and can be used for standard ICA as well as for overcomplete ICA (i.e. more source signals than observed signals (mixtures)).

EPrint Type:Article
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Theory & Algorithms
ID Code:2020
Deposited By:Frank Meinecke
Deposited On:15 January 2006