Title: Statistics of feature extraction by topographic independent component analysis from natural images
Authors: Mutihac, Radu ×
Van Hulle, Marc #
Issue Date: 2003
Host Document: Proceedings vol:CD rom pages:451-289
Conference: International Conference on Electronics, Control and Signal Processing (World Scientific and Engineering Academy and Society) edition:2 location:Singapore date:7-12 December 2002
Abstract: Our contribution highlights the statistical properties and biological interpretation of the basis
vectors (filters) that result from applying topographic independent component analysis (ICA) to feature
extraction from patches of natural images. The consistency of the feature sets obtained from various
collections of natural image data sets applying topographical ICA (TICA) supports the opinion that the
statistical properties of the environmental stimuli enforce a process according to some optimization criterion,
which provides a good computational model for the response properties of sensory neurons. However, the
basis vector set differs statistically meaningful from one image collection to the other, making the ICA
decomposition of natural images unsuitable for a novel approach to image compression.
Publication status: published
KU Leuven publication type: IC
Appears in Collections:Research Group Neurophysiology
× corresponding author
# (joint) last author

Files in This Item:

There are no files associated with this item.

Request a copy


All items in Lirias are protected by copyright, with all rights reserved.