Compressed sensing and multistatic SAR

We demonstrate that the remarkable advantages of compressed sensing remain in force when the information operator is constrained to obey the physical rules of a multistatic SAR measurement. The design guidelines of the SAR information operator for ℓ2 reconstructions is compared to those provided for generic ℓ1 reconstructions. We report little or no degradation in compression performance when using an information operator obeying SAR sampling constraints. Simulations for a Shepp-Logan image show an image is faithfully reconstructed when the number of measurements is about a third of the number of image pixels, using a minimum total-variation technique. We observed high sensitivity in performance and algorithm convergence to small perturbations in the measurement vectors.