Scale Mixture of Gaussians Modelling of Polarimetric SAR Data
- 1 June 2006
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 18-21
Abstract
This paper discusses a multivariate, non-Gaussian parametric modelling technique to analyse polarimetric SAR data. We investigate a simple class of multivariate non-Gaussian distributions, the 'scale mixture of Gaussians', and assess its "Goodness-of-fit" to the radar data. Four models are analysed and various characteristics of the models are interpreted, together with practical considerations with regard to parameter estimation. We observe that SAR data is often not Gaussian in distribution, being more highly peaked at zero and falling off more slowly than the Gaussian. It is shown that a single 'flexible' model is sufficient to capture the statistics of the SAR data, leading to a feature set of the modelled parameters. Image classification is then studied by means of the modelled data and compared with an existing land cover mapKeywords
This publication has 3 references indexed in Scilit:
- The Rician inverse Gaussian distribution: a new model for non-Rayleigh signal amplitude statisticsIEEE Transactions on Image Processing, 2005
- K-Distribution and Polarimetric Terrain Radar ClutterJournal of Electromagnetic Waves and Applications, 1989
- A model for non-Rayleigh sea echoIEEE Transactions on Antennas and Propagation, 1976