Speckle reduction in SAR imagery by Lp normed filtering

The presence of speckle in SAR images increases the probability of false alarms in image prescreening algorithms designed to automatically detect small targets. In order to reduce the level of speckle in SAR images a wide variety of image processing techniques have been developed. Speckle reduction has been accomplished via low pass filtering, median filtering, adaptive filtering, wavelet based filtering, and various iterative techniques such as the gamma filter. For this paper we develop an L_p normed filter structure. The filter weights are adaptive, based upon a conventional multiplicative noise model of the speckle phenomena in SAR imagery. The adaptive filter for the L_p, p=2 case, is equivalent to the "Frost" filter, and is based upon second order statistics of the multiplicative noise and assumed image model. The filter is tested against SAR data and is shown to reduce the number of false alarms as compared to several other speckle reduction filters.