Reliably estimating the speckle noise from SAR data

A fully unsupervised method is proposed to assess the variance and the spatial correlation coefficients of speckle noise in observed SAR images. The former is obtained as regression coefficient of the local standard deviation to local the mean, the latter from the scatterplot of local unity-lag covariance to local variance; both must be calculated on constant-signal areas. In order to overcome the drawback of manually identifying homogeneous areas, an automatic procedure has been developed, based on considerations that such areas tend to produce clusters of points which are aligned along the regression straight line. Results on simulated speckled images show an impressive accuracy. On true SAR images the authors note that the method is capable to carefully reject textured regions, in which the speckle may be not fully developed and the variance of the signal is not negligible.