Enhanced Goldstein Filter for Interferometric Phase Denoising Using 2-D Variational Mode Decomposition

Abstract

Denoising of interferograms is a vital step in the processing of synthetic aperture radar (InSAR) data. The primary goal is to filter the noise to the extent possible while retaining the fringes of the interferograms. Among the widely available classes of filters, the frequency-domain filters are still being used, owing to their robustness and generalizability to varying phase noise characteristics. This article deals with an enhancement to the well-known frequency-domain filter, i.e., the Goldstein filter, which is basically a phase filtering algorithm for interferometric products. The proposed extension to the Goldstein filter deals with deriving the tuning parameter based on the spatial frequency modes. This is achieved by using the mode-level characteristics rendered by the 2-D version of variational mode decomposition (2D-VMD) on the interferograms under test. The results of simulation and real interferogram data show that the proposed approach reduces the noise levels while minimizing the loss of signal.