Dictionary learning-based approach for SAR image reconstruction

Recently there has been growing interest on the study of sparse representation-based SAR imaging with the assumption that the underlying SAR scenes exhibit sparsity with respect to SAR image features such as point scatterers and edges of smooth regions. Since the phase of the SAR reflectivity is random, these methods have been employed to magnitude of the complex valued SAR data. Pre-defined overcomplete dictionaries are used for these methods. In this paper we propose a dictionary learning-based SAR image reconstruction method. Our proposed model learns the dictionary from training set of images. We validate our method on synthetic SAR image.