Bayesian dynamic experiment design regularization framework for high-resolution radar/SAR imaging

The problem of high-resolution array radar/SAR imaging is stated and treated as a nonlinear ill-posed inverse problem of nonparametric estimation of the power spatial spectrum pattern (SSP) of the wavefield scattered from the remotely sensed scene observed through a linear signal formation operator. We develop the radar/SAR adapted imaging framework that combines the “model-based” Bayesian experiment design (BED) high-resolution spatial spectral estimation strategy with the “model-free” dynamic variational analysis (VA)-based image enhancement approach that incorporates projections onto convex solution sets into the SSP reconstruction procedures to sharpen the image edge map, enforce the robustness and speed-up the convergence. We also show how the proposed unified BED-VA framework may be considered as a dynamic VA-based generalization of the robust MVDR, APES and other high-resolution nonparametric radar imaging techniques. A family of iterative BED-VA-related algorithms with substantially enhanced resolution performances is constructed and their effectiveness is illustrated via numerical simulations.