A Novel Non-Local Denoising Filter Based on Multibaseline InSAR

Abstract

Denoising filtering is one of the most critical steps in interferometric synthetic aperture radar (InSAR) data processing. There are many denoising filtering algorithms, which are suitable for different specific scenarios. However, there is a contradiction between detail retaining and noise reduction at the same time, especially for areas with large terrain fluctuations. In order to solve such a contradiction, an improved nonlocal denoising filtering algorithm based on the multibaseline InSAR is proposed in this article. Based on the relationship between interferometric phases with the multiple baselines, we calculated the joint probability by a nonlocal probability density function (PDF) to effectively preserve fringes, especially for the interferogram with a large baseline. Combined with the PDF obtained by machine learning, we got more satisfactory results with better continuity of fringes and the details of the interferograms as well as maximizing noise reduction.