Plug-and-Play ADMM for Sparse ISAR Imaging

Under the condition of sparse aperture, sparse signal recovery (SSR) theory is generally utilized to obtain radar images. However, the existing SSR methods require heavy computation with a lot of iterations. Recently emerging plug-and-play (PnP) prior technology is a flexible framework that combines the forward imaging model with the prior model. Inspired by this, we proposed a novel PnP framework for sparse ISAR image reconstruction, which utilized a complex-valued convolutional neural network to denoise ISAR images and learn prior information. Thus, the alternating direction method of multipliers (ADMM) based imaging algorithm combined with the trained denoising network can take advantage of learning-based methods and model-based methods. Simulation and experimental results show that the proposed PnP-ADMM imaging algorithm can obtain high imaging quality results while maintaining efficient computation compared with the state-of-the-art sparse ISAR imaging method.