Detecting variable source areas from temporal radar imagery using advanced image enhancement techniques

Recently, N. E. C. Verhoest et al. (1998) showed that it is possible to map variable source areas in a catchment using a principal component analysis. This technique, based on a temporal series of images, revealed the spatial soil moisture patterns from the vegetation and topographic effects introduced in a synthetic aperture radar (SAR) image. However, the obtained image is still corrupted with noise, which is partially related to the speckle observed within a SAR image. In order to get a noiseless image, which is more appropriate for hydrological modelling schemes, the authors apply a recently developed wavelet-based image denoising technique, A. Pizurica et al. (1999). The main advantage of this filtering technique is that it preserves the spatial patterns and observed edges, while it increases the signal to noise ratio significantly. The suitability of this denoising algorithm is investigated by comparing the hydrologic information included in these visually well-appearing images with the results obtained for their non-filtered counterparts