Sparse Representation Based SAR Imaging Using Combined Dictionary

Abstract

Sparse representation (SR)-based SAR imaging has shown its superior capability in high-resolution image formation. For SR-based SAR imaging task, a key challenge is how to choose a proper dictionary that can effectively represent the magnitude of the complex-valued scattering field. In this paper, we present a combined dictionary that simultaneously enhances multiple types of scattering mechanism. Trained by different kinds of SAR image patches with either strong point scatterers or smooth regions, the dictionary can represent both point-scattering and spatially distributed scenes sparsely. Finally, the SAR image is obtained by solving a joint optimization problem over the combined representation of the magnitude and phase of the observed scene.