SAR interferogram phase filtering based on the Von Mises distribution

We propose the use of the Von Mises distribution for circular data to model interferometric phase images. Based on the kappa parameter of the Von Mises distribution a locally adaptive phase filter is developed. A two-sided confidence interval on the kappa values is constructed, and only phase measurements where kappa values fall inside this interval are input to the filter. We additionally put more importance to spatially closer pixels than to pixels further away from the pixel in question. The latter idea is achieved by narrowing the confidence interval depending on the distance from the central pixel.