Adaptive anisotropic diffusion for speckle filtering in SAS imagery

Thanks to their high resolution, the images provided by synthetic aperture sonar (SAS) are of great interest for the detection and classification of underwater mines lying on the sea bed or buried in the sea floor. As with any acoustic imaging systems, SAS data are highly corrupted by a multiplicative noise, called the speckle, which can be very disturbing for the interpretation and the automatic analysis of the images. In order to reduce this noise and smooth the data, the properties of an anisotropic diffusion filter (ADF) are interesting. In this paper, we propose to adapt the setting of this filter parameters thanks the knowledge provided by an other process. This allows to smooth the speckle noise in the background, while the regions likely to contain any object of interest are preserved.