Wishart distribution based level set method for polarimetric SAR image segmentation

We develop a level set segmentation method for a wide range of SAR data from single channel intensity data to multifrequency and/or multitemporal polarimetric data in this study. By modeling of the minimization functional for segmentation and the complex Wishart distribution for polarimetric SAR data representation, we propose a model consisting of three parts, an original `fitting´ term derived from the maximum a posteriori (MAP) estimator, a classical curve length term that drives the zero level set toward the object boundaries, and the region term of interest. A CFAR Polarimetric SAR edge detector is used as the edge indicator, and also a penalty term is employed to completely eliminate the costly re-initialization procedure during the traditional level set evolution. The proposed method of the two-phase segmentation is extended to a multiphase case. Good results are obtained by using both simulated polarimetric data and two NASA-JPL L-band polarimetric SAR images.