Super-Resolution Power of Deep Unfolding Networks to TomoSAR 3D Imaging

Deep unfolding networks have been proved to be able to achieve super-resolution (SR). Different from sparsity regularization methods, deep unfolding networks can speed up the imaging process and own good robustness against noise. However, the optimal network design and maximum SR ability of deep unfolding networks have not been quantified. In this paper, the optimal number of layers and the SR extremity of the network are confirmed by simulation. And some sparsity regularization methods are compared with the proposed method. Simulation results validate the superiority of the proposed method in terms of SR imaging and accuracy.