Segmentation of lakes from the local background on the surface of Titan using Cassini SAR images

Synthetic Aperture Radar (SAR) images of Titan, the largest satellite of Saturn, reveal quasi-circular to complex features which are interpreted to be liquid hydrocarbon lakes. One of the major problems hampering the derivation of meaningful texture information from SAR imagery is the speckle noise. It overlays real structures and causes gray value variations even in homogeneous parts of the image. A filtering technique is applied to obtain the restored SAR images. Our method is based on probabilistic methods and regards an image as a random element drawn from a prespecified set of possible images. The TSPR (Total Sum Preserving Regularization) filter used here is based on a membrane model Markov random field approximation with a Gaussian conditional probability density function optimized by a synchronous local iterative method. The despeckle filter can be used as intermediate stage for the extraction of meaningful regions that correspond to structural units in the scene or distinguish objects of interest like lakes.