LAUSR

ABSTRACT Phase filtering is an important processing step in interferometric synthetic aperture radar (InSAR) measurement. It is a difficult task to achieve noise reduction and fringe preservation simultaneously. To solve the dilemma, a new noise reduction scheme for InSAR phase images is introduced based on total generalized variation (TGV). The TGV regularization has been mathematically proven to be able to preserve edges while still imposing high-order smoothness in homogeneous regions. Combined with the characteristics of phase image, a spatially adapted TGV model is used as a regularizer in phase filtering. The regularization parameters are adaptively selected according to the coherence and local phase image features. The presented model is realized in the complex domain in order to avoid the effect of phase wrapping. Experiment results from simulated and real data sets demonstrate the effectiveness of this algorithm.