Hyperspectral data selection from mutual information between image bands

Abstract

This work presents a band selection method for multi and hyperspectral images using correlation among bands based on mutual information measures. The relationship among bands are represented by means of the transinformation matrix. A process based on a Deterministic Annealing optimization is applied to the transinformation matrix in order to obtain a reduction of this matrix looking for the image bands as less uncorrelated as possible between them. Some experiments are presented to show the effectiveness of the bands selected from the point of view of pixel classification. © Springer-Verlag Berlin Heidelberg 2006.