Robust Speckle Reducing Anisotropic Diffusion

Inhalt:
Speckle Filter performance strongly depends on speckle model and heterogeneity measurement. Classical singlestage speckle filters fails to retain image structures on iteration, but iteration is necessary for effective smoothing of speckle. In this paper we propose a heterogeneity measuring technique that enables to apply traditional filters iteratively. We also propose a new robust anisotropic diffusion (AD) based speckle filter (ROSRAD), that shown to perform better than the existing speckle reducing AD filters like SRAD and DPAD. Performance of ROSRAD is tested on simulated and real SAR images and compared with other AD filters.