Abstract: In order to obtain a more accurate topographic model, anoise filtering step must be performed before theunwrapping of phases. We propose an iterative processthat involves a filter in the spatial domain and amulti-resolution phase unwrapping method. The approachis based on the generation of an approximate phasemodel, which is iteratively refined. In the aim ofpreserving fine details in the interferogram that aredirectly related to the ground topography, an edge-preserving smoothing method has been applied. The localspatial content of the phase image is exploited inorder to determine for each pixel the best matchingshape of the filtering mask, searching in a set ofdifferent neighborhood systems. Then, a non-linearadaptive filtering function, based on the localestimation of noise and signal standard deviation, isadopted. The results obtained by processing severalnoisy simulated and real interferometrical images withthe described method show clearly how the combinationof a detail preserving adaptive filter with a noiserobust phase unwrapping approach appears as a goodmeans for reducing the influence of distributed noiseand low coherence areas on the determination of adigital ground elevation model.!12
展开▼
