Exploitation of the additive component of the polarimetric noise model for speckle filtering

Abstract

Ratio filters for speckle noise reduction in SAR imagery are recursive filters where the image structure is iteratively recovered from an initial oversmoothed image. We show that the MBPolSAR filter could be interepreted as a ratio filter applied to the off-diagonal terms of the covariance/coherency matix. From this observation, we propose a new polarimetric ratio filter allowing us to recover the image structure from all the terms of the covariance matrix. In addition, we briefly look at how the additive noise component could also be exploited for the image structure extraction. Filtering results on both simulated and real PolSAR images are shown.