SAR image despeckling using the multiscale edge representation

The authors compare a despeckling technique which employs a wavelet algorithm, the Multiscale Edge Representation (MER), to several adaptive spatial filters. The authors have seen positive- results in using the MER as part of a speckle reduction technique and include a comparison of the effects of speckle edge removal to speckle edge reduction. Speckle is modeled as multiplicative Gaussian noise. Simulated images are used to quantitatively evaluate despeckling methods in a number of measures including reduction of the image standard deviation in homogeneous regions, maximization of the signal to mean squared error and preservation of the image mean and histogram shape. The authors´ findings indicate that the MER is an effective, tunable, edge preserving despeckling technique