Fractal dimensions segmentation of synthetic aperture radar images

Presents an algorithm for segmenting an image into regions of similarity based on computing the fractal dimension-Fractal Dimension Segmentation (FDS). Unlike other work which has been published in this area, the technique discussed here is based exclusively on the application of a fast Fourier transform (FFT). Coupled with a new least squares approach for estimating the fractal dimension of either a signal or image, the application of a FFT provides the potential for acquiring a near real time facility for segmenting an image using FDS by implementing suitable hardware. In the paper, the algorithm is demonstrated using Synthetic Aperture Radar images. SAR provides fully coherent images which vary considerably in dynamic range and texture. Also, SAR images are separable (i.e. to a good approximation, possess a separable point spread function) which allows FDS to be undertaken on a single-by-signal basis. Using this approach, it is shown how the fractal dimension of a SAR image to be viewed quickly and effectively, and in particular, it is demonstrated that the algorithm may be of value for target detection.