BM3D vector approximate message passing for radar coded-aperture imaging

Terahertz radar coded-aperture imaging technology can achieve high-resolution, forward-looking and staring imaging by producing spatiotemporal independent signals in the imaging area. Common Compressive Sensing (CS) algorithms are efficient in the point target imaging, whereas it becomes invalid for natural targets without sparse prior. In contrast, the denoising-based vector approximate message passing (D-VAMP) methodology, recently proposed by Rangan, allows one to plug in sophisticated denoisers like BM3D into the VAMP algorithm to achieve state-of-the-art compressive image recovery. In this paper, we show that the benefits of D-VAMP for natural scene through numerical simulations.