福田 盛介

A classification scheme of land covers in SAR images using wavelet-based textural information is presented. The proposed scheme treats multifrequency polarimetric data. Extraction of texure features, polarization selection, feature reduction, and so on, are discussed. The scheme has been successfully applied to the Flevoland site.

This paper proposes new techniques of multiresolution analysis and processing of synthetic aperture radar (SAR) images using wavelets. Here, two works are presented; one is speckle reduction by reducing the amplitude of the detail images in wavelet subspaces, while preserving edges by releasing the amplitude reduction around edges, and the other is estimation of fractal dimension which is featuring 'roughness' of the 2-D surface, by wavelet-based multiresolution analysis combined with the fractal Brownian motion (fBm). Both methods have been satisfactorily applied to real SAR images, and are expected to be useful for SAR remote sensing.