Oil-spills detection in SAR images by fractal dimension estimation

The paper describes a multi-resolution algorithm based on fractal geometry for texture analysis and detection of oil spills in SAR images. The multi-resolution approach reduces the problems of speckle and sea clutter and preserves subtle variations of oil slicks. The use of fractal dimension as a feature for classification improves the oil spill detection, since enhances texture discrimination. The proposed approach computes the fractal dimension from power spectra: the ratio between powers at different scales is straightforwardly related to fractal dimension. The proposed method to compute a fractal signature is based on the multi-scale image decomposition obtained by the Normalized Laplacian Pyramid which provides a reliable estimation of the fractal dimension even in the presence of speckle.