Feature Enhancement of Interferometric Synthetic Aperture Radar Image Formation Using Sparse Bayesian Learning in Joint Sparsity Approach

Abstract

A novel sparse Bayesian learning approach with a joint sparsity model is proposed for Interferometric Synthetic Aperture Radar (InSAR) image formation to realize the feature enhancements of interferometric phase denoising and speckle reduction. Using Bayesian rules, sparse image formation is achieved using a hierarchical statistical model. In particular, structured sparsity with joint channels is imposed on the InSAR images. During sparse imaging, an Expectation-Maximization (EM) method is employed for image formation and hyper-parameter estimation. Using joint sparsity statistics, the performance of the noise reduction on the magnitude and phase of InSAR images can be improved. Finally, experimental analysis is performed using simulated and measured data to confirm the effectiveness of the proposed algorithm.