Auto-focus for under-sampled synthetic aperture radar

We investigate the effects of phase errors on undersampled synthetic aperture radar (SAR) systems. We show that the standard methods of auto-focus, which are used as a postprocessing step, are typically not suitable. Instead of applying auto-focus as a post-processor we propose using a stable algorithm, which is based on algorithms from the dictionary learning literature, that corrects phase errors during the reconstruction and is found empirically to recover sparse SAR images.