Reducing the radiometric terrain effect in SAR imagery by means of principal components analysis

SAR images of hilly and mountainous regions are dominated by the terrain-induced backscatter variation. To apply automatic classification methods, e.g. for land cover discrimination, this influence has to be reduced in order to get meaningful results. The application of backscatter models based on the local incidence angle is a complex, so far not operational task, especially: in heterogeneous landscapes with varying land uses which are not known a priori. Principal Components Analysis (PCA) has been an important methodologic tool for the analysis of image time series during the last decades, nevertheless most applications were performed on optical imagery. Henebry (1997) discussed the advantages of PCA for land cover segmentation from SAR image series. The following article focuses on the use of PCA for the reduction of radiometric terrain effects in SAR data