SAR Image Classification by Wavelets

The aim of this work is to propose a method for classifying Synthetic Aperture Radar (SAR) images based on a multiresolution representation of the images obtained by Wavelet Transform (WT). The WT offers an efficient and nonredundant tool f o r analyzing the image at different scales and, equivalently, at different spatial frequency bands and orientations. The analysis of the energy content at different scales and orientations can be exploited to discriminate areas with stronger texture variability, such as urban areas, from less structured regions, such as cultivated areas. The effect of the speckle is mitigated by a noncoherent smoothing in the the wavelet domain. The wavelet is also used as a segmentation tool. The proposed approach is tested on real SAR images as well as on simulated images to quantify the percentage of correct classification.