SAR image segmentation through B-spline deformable contours and fractal dimension

Synthetic Aperture Radar (SAR) images are usually corrupted by a signal-dependent non-additive noise called speckle. This makes difficult the segmentation, object identification, and feature extraction within this kind of images. In this work we propose the combination of local fractal estimation and B-Spline based active contours as a solution for the boundary extraction problem in SAR images. After a supervised initialization (the specification of a an initial curve laying com-pletely within the region of interest), the algorithm searches the control points (vertices) of a B-Spline curve that fits the boundary of the region to be segmented. The vertices of the curve are found by a local estimation of the fractal dimension in the surrounding. Fractal dimension provides a good local roughens and statistical correlation estimation. Box-counting measurement of the fractal dimension is widely acknowledged to be the most adequate in terms of robustness and com-putational requirements. Box counting algorithms are based on a statistical analysis of the brightness distribution of the pixels in a surrounding of varying sizes. A power law can be established between the surrounding size and the amount of pixels with a given brightness profile, and then an adequate assessment of the local fractal dimension can be per-formed. The proposed algorithm is systematically tested on synthetic and real SAR images, and both the accuracy and the performance of our proposal are assessed.