A Simple Feature Extraction for High Dimensional Image Representations
Christian Savu-Krohn and Peter Auer
In: Subspace, Latent Structure and Feature Selection techniques: Statistical and Optimisation perspectives, 23-25 Feb 2005, Bohinj, Slovenia.

Abstract

We investigate a method to find local clusters in low dimensional subspaces of high dimensional data, e.g. in high dimensional image descriptions. Using cluster centers instead of the full set of data will speed up the performance of learning algorithms for object recognition, and migh t also improve performance because overfitting is avoided. Usingthe Graz01 database, our method outperforms the current standard method for feature extraction from high dimensional image respresentations.

[Edit]