Plug-and-play algorithms for single-pixel imaging

Single-pixel imaging (SPI) captures images using a photodiode, which has advantages of low cost and high signal-to-noise ratio compared with sensor arrays. When combining with compressed sensing (CS), images can be recovered from fewer measurements in spite of the ill-posed problem. The most commonly used SPI reconstruction algorithms are model-based methods, which usually utilize the hand-crafted prior and limit image quality in the case of under-sampling. In this paper, we propose a novel compressive imaging approach for SPI based on Plug-and-Play (PnP) algorithms. It well employs deep denoiser prior and combines with model-based algorithm advantage. Extensive results on both simulation and real data show that the proposed PnP-based SPI reconstruction method outperforms state-of-the-art SPI algorithms and realizes higher quality image reconstruction.