Clustering by deterministic annealing andWishart based distance measures for fully-polarimetric SAR-data

Conference: EUSAR 2008 - 7th European Conference on Synthetic Aperture Radar
06/02/2008 - 06/05/2008 at Friedrichshafen, Germany

Proceedings: EUSAR 2008

Pages: 4Language: englishTyp: PDF

Personal VDE Members are entitled to a 10% discount on this title

Authors:
Hänsch, Ronny; Jäger, Marc; Hellwich, Olaf (Berlin University of Technology (TUB), Department of Electrical Engineering and Computer Science Computer Vision and Remote Sensing, Sekr. FR 3-1, Franklinstr. 28/29, D-10587 Berlin)

Abstract:
The basic assumption of clustering is that the observed data embodies a specific structure. This structure is represented by groups of data points, which are relativly isolated within the feature space. The goal of clustering is to find and represent this dissimilar groups of similar elements or, in other words, to subdivide the data space into a number of partitions. In this paper a spectral clustering based on deterministic annealing is used to obtain image segments with similar spectral properties. The results are compared with other established clustering methods for polarimetric SAR data and advantages of this approach are investigated.