Fast Marginalized Sparse Bayesian Learning for 3-D Interferometric ISAR Image Formation Via Super-Resolution ISAR Imaging

Compressed sensing (CS) has been successfully applied to inverse synthetic aperture radar (ISAR) imaging of moving targets. In existing CS-based ISAR imaging algorithms, Laplace distribution is widely adopted to enforce sparseness on signal recovery. However, this kind of CS method using Laplace prior encounters the problems of determining optimum regularization factor and heavy computation load. In this paper, a fast marginalized sparse Bayesian learning (MSBL) method is proposed for three-dimensional (3-D) interferometric super-resolution ISAR imaging. After deriving the target sparsity-driven imaging model, a fast MSBL approach is applied to obtain super-resolution ISAR image, and then a high-quality 3-D view of the target is achieved via the interferometry technique using an ISAR imagery pair. Experiments on simulated and real data are provided to validate the effectiveness of the proposed method.