Plug-and-Play Regularization on Magnitude with Deep Priors for 3D Near-Field MIMO Imaging

TL;DR

This work addresses 3D near-field MIMO radar imaging by reconstructing the complex-valued scene reflectivity while enforcing regularization on its magnitude. It derives a closed-form proximal mapping for magnitude-based regularizers within an ADMM-based plug-and-play framework and introduces a learned 3D denoiser as a data-driven prior, enabling efficient, high-quality reconstructions from highly compressive and noisy measurements. The approach is validated on simulated and experimental data, showing state-of-the-art performance and fast runtimes compared to direct inversion and sparsity-based methods, and it generalizes to other radar imaging problems such as SAR. Overall, the method provides a unified, adaptable framework for magnitude-domain regularization with deep priors, suitable for practical 3D radar image formation tasks and potentially other complex-valued inverse problems.

Abstract

Near-field radar imaging systems are used in a wide range of applications such as concealed weapon detection and medical diagnosis. In this paper, we consider the problem of reconstructing the three-dimensional (3D) complex-valued reflectivity distribution of the near-field scene by enforcing regularization on its magnitude. We solve this inverse problem by using the alternating direction method of multipliers (ADMM) framework. For this, we provide a general expression for the proximal mapping associated with such regularization functionals. This equivalently corresponds to the solution of a complex-valued denoising problem which involves regularization on the magnitude. By utilizing this expression, we develop a novel and efficient plug-and-play (PnP) reconstruction method that consists of simple update steps. Due to the success of data-adaptive deep priors in imaging, we also train a 3D deep denoiser to exploit within the developed PnP framework. The effectiveness of the developed approach is demonstrated for multiple-input multiple-output (MIMO) imaging under various compressive and noisy observation scenarios using both simulated and experimental data. The performance is also compared with the commonly used direct inversion and sparsity-based reconstruction approaches. The results demonstrate that the developed technique not only provides state-of-the-art performance for 3D real-world targets, but also enables fast computation. Our approach provides a unified general framework to effectively handle arbitrary regularization on the magnitude of a complex-valued unknown and is equally applicable to other radar image formation problems (including SAR).

Figures (11)

• Figure 1: Schematic view of a near-field MIMO radar imaging system.
• Figure 1: Denoising performance of different methods; (a) average test PSNR with respect to noise standard deviation $\sigma_\nu$, (b) ground truth magnitudes of the sample test image, (c) noisy input magnitudes at $\sigma_\nu=0.2$, (d)-(f) denoised outputs corresponding to $\ell_1$, $TV$ and deep-prior based denoising and the respective PSNRs (dB).
• Figure 2: Developed PnP Method for Complex-valued Reconstruction with Regularization on Magnitude.
• Figure 2: Imaged revolver and its reconstructions using 11 frequency steps with experimental data. Images have a 2mm resolution.
• Figure 3: Network architecture of the proposed 3D deep denoiser. “C”, “B”, “R”, “Max. Pool.” and “T.Conv.” represent 3D convolution, batch normalization, ReLU activation, max-pooling operation, and transposed convolution, respectively. The number of output channels is denoted inside parentheses.