Space-Adaptive coherence estimation
Andrea Monti Guarnieri(1) and Marco Arrigoni(2)
Politecnico di Milano,
piazza Leonardo da Vinci, 32,
(2) Aresys s.r.l. - a POLIMI spin off company, via Garofalo, 39, 20133, Italy
Accurate, un-biased coherence estimation is a key issue in interferometric applications.
The performances of classic coherence estimators, based upon a fixed window, are strongly dependent on estimation window size and shape. Small windows lead to a coherence bias, whose magnitude depends on the coherence itself, whereas large windows lead to poor spatial resolution. Moreover, coherence estimate performances are highly affected by the non-stationarities of the amplitudes. We propose an adaptive algorithm that estimates the interferometric coherence by exploiting a window whose size and shape adapts locally, in a region growing fashion.
The number of samples in the window is chosen according to the expected
coherence, based on an a-priori estimate, and the interferogram resolution, that in turn depends upon the local slope. The shape of each window is chosen in order to cover a connected set of pixels with uniform backscatter. The resulting coherence maps attain high spatial resolution and good accuracy at the same time, giving low bias and preservation of the small scale features.
An evaluation of the algorithm performances and a comparison with traditional methods are also included.