Noise and speckle reduction in synthetic aperture radar imagery by nonparametric Wiener filtering

We present a Wiener filter that is especially suitable for speckle and nois e reduction in multilook synthetic aperture radar (SAR) imagery. The propos ed filter is nonparametric, not being based on parametrized analytical mode ls of signal statistics. instead, the Wiener-Hopf equation is expressed ent irely in terms of observed signal statistics, with no reference to the poss ibly unobservable pure signal and noise. This Wiener filter is simple in co ncept and implementation, exactly minimum mean-square error, and directly a pplicable to signal-dependent and multiplicative noise. We demonstrate the filtering of a genuine two-look SAR image and show how a nonnegatively cons trained version of the filter substantially reduces ringing. (C) 2000 Optic al Society of America OCIS codes: 100.2000; 100.2980; 100.3020; 280.6730.