One of the major factors plaguing the performance of synthetic aperture radar (SAR) imagery is the signal-dependent, speckle noise. Grainy in appearance, it is due to the phase fluctuations of the electromagnetic returned signals. Since the inherent spatial-correlation characteristics of speckle in SAR images are not embedded in the multiplicative models for speckle noise, a new mathematical framework for modeling speckled imagery is introduced. It is based on embedding the spatial correlation properties of speckled imagery, obtained from statistical optics, into a Markov-random-field (MRF) framework. The model is then used to perform speckle-noise reduction through the utilization of a global energy-minimization algorithm, which consists of simulated annealing in conjunction with the Metropolis sampler algorithm. A comparative study using both simulations and real SAR images indicates that the proposed approach performs better compared to filtering techniques such as the Gamma Map, the modified-Lee and the enhanced-Frost algorithms. This success is attributable to the ability of the proposed model to capture the physical spatial statistics of speckle within the confines of a MRF framework.
Lankoande, Ousseini; Majeed M. Hayat; and Bal Santhanam. "Speckle noise modeling and reduction of SAR Images based Markov random fields." (2005). https://digitalrepository.unm.edu/ece_rpts/16