Accurate Pixel-Based Noise Estimation for InSAR Interferograms

Distortions in Interferograms is unavoidable in interferometric Synthetic Aperture Radar (InSAR) capture and processing. A large part of these distortions manifest as noise in the phase of the InSAR interferograms. In this paper, we propose a convolutional neural network based framework to detect these phase distortions. The neural net is trained with data from a custom InSAR simulator. We demonstrate that the proposed system can accurately detect noise. We also show applications of the proposed noise detector as a coherence estimator and a noise masking mechanism.