Passive compressive ghost imaging with low-rank optimization

High-quality passive ghost imaging with a low sampling rate can effectively improve the channel capacity utilization of optical imaging systems, but the current ghost imaging methods generally suffer from the problem of insufficient channel capacity utilization. We take advantage of the property that the matrix composed of nonlocal similar blocks of an image has a low rank, combine it as a priori information with the compressed sampling model of ghost imaging, establish a ghost imaging model based on low-rank optimization, and propose an alternating minimization solution method for this model. According to the simulation and experimental verification of the ghost imaging camera, our proposed ghost imaging method outperforms similar imaging methods in PSNR, SSIM, as well as other objective evaluation indexes. Among them, for the reconstruction of a typical scene at a sampling rate of 6.25 %, our method outperforms peer methods by 20.6 % for PSNR and 68.7 % for SSIM. Furthermore, we analyze the confidence intervals and uncertainties of different ghost imaging methods in the process of optical field decoding, and the experimental results show that our imaging method has a higher confidence level.