Optimum detection and segmentation of oil-slicks using polarimetric SAR data

The paper is concerned with techniques for optimising the detection and definition of slick boundaries on the sea surface using polarimetric imagery. In principle, the full polarimetric return should provide more information than is available in a single polarisation. The authors compare the performance of a set of different polarisation measures applied to the detection of slicks. Annealed segmentation of these measures is then employed to detect and define their boundaries. Theoretical predictions are derived for the probability of detection using conventional polarisation measures, including the intensity in a single polarisation and the maximum eigenvalue and span measures for more than one polarisation channel. The authors also propose two implementations of a maximum likelihood polarisation discriminant and demonstrate that these yield significant improvement in slick detection and boundary definition.