Non-linear Filtering And Edge Detection In Speckled Radar Images
J.-M. Boucher, A. Hillion · 2005
A new non-linear filter reducing the speckle noise on radar images is examined.This filter ,which is better than the classical ones according to the mean-square error criterion.depends on a real parameter a.For a equal to zero .the non-linear filter takes a logarithmic structure.and for a equal to one,one gets the classical linear filter(LEE filter).A comparison between these two methods is made in the particular case of a signal taking only two values with a lognormal density of probability for the noise..An adaptive algorithm,involving both filters,can be applied to radar images,showing that edges can be detected by comparing the computed mean square errors of these filters.