The Potential of Multi-Agent Consensus Equilibrium for Synthetic Aperture Radar Processing

Abstract:
Multi-agent consensus equilibrium(MACE) mechanism is a generalization of popular used PnP-ADMM method in computational imaging. We propose a novel SAR processing framework based onMACE mechanism, named byMACE-SAR. The proposed framework offers the possibility to unify various SAR processing agents of both data-driven learned priors and physics-based regularization priors, enabling the integration of diverse functions and enhance the image reconstruction quality. We exemplifies how the MACE-SAR framework can be customized to combine SAR image formation, feature enhancement, image denoising, target detection, etc. The effectiveness of the framework is demonstrated through experiments conducted on simulated and Gaofen-3 satellite real data.