STRUCTURE DETECTION AND STATISTICAL ADAPTIVE SPECKLE FILTERING IN SARIMAGES

Current speckle filters attempt to restore the radar reflectivity usin g only the multiplicative speckle noise assumption. The best known fil ters, namely the Frost, Lec or Kuan filters arc adaptive filters based on the local statistics, computed in a fixed square window. In this w ay, the speckle is reduced as a function of the heterogcneity measured by the local coefficient of variation. When the radar reflectivity un dergoes significant variations due to the presence of strong scatterer s or structural features (edges or lines) in the processing window, su ch speckle filtering is less effective. In this paper it is shown that the filtering process can be controlled both by the coefficient of va riation and by various geometrical ratio detectors. Through shape adap tive windowing, these detectors allow the use of large window sizes fo r better speckle reduction while preserving spatial resolution and str uctural features. The backscattered intensity is modelled as K-distrib uted within speckled targets, and the filter uses a Bayesian approach which allows an explicit use of the multiplicative noise model and the radar reflectivity distribution.