Synthetic aperture radar segmentation using wavelets and fractals

It is shown that fractal dimension estimates and Gabor wavelet coefficients are valid features of segmenting high-resolution polarimetric synthetic aperture radar imagery. Results of training a radial basis function neural network using fractal dimension features, Gabor wavelet coefficients, and a combination of both fractal and Gabor wavelet features are presented. Current research into combining these two techniques both theoretically and empirically is presented. One-foot resolution polarimetric synthetic aperture radar imagery is successfully segmented into culture, tree, field, and shadow regions.<<ETX>>