Plug-and-Play Reconstruction with 3D Deep Prior for Complex-Valued Near-Field MIMO Imaging

Near-field radar imaging systems are used in a wide range of applications, such as medical diagnosis, through-wall imaging, concealed weapon detection, and nondestructive evaluation. In this paper, we consider the inverse problem of reconstructing the three-dimensional (3D) complex-valued reflectivity distribution of the near-field scene from the sparse multiple-input multiple-output (MIMO) array measurements. Using the alternating direction method of multipliers (ADMM) framework, we formulate this problem by exploiting regularization on the magnitude of the complex-valued reflectivity distribution. We then provide a general expression for the proximal mapping associated with such regularization functionals operating on the magnitude of the complex-valued unknown. By utilizing this expression, we develop a computationally efficient plug-and-play reconstruction method that involves simple update steps both with analytical and deep priors. We illustrate the reconstruction performance of our approach with a 3D deep prior on a synthetic dataset. We also compare the result with the classical back-projection method and magnitude-total variation. Our results demonstrate that significant performance improvement can be achieved with learned 3D priors.