Strongly robust computational ghost imaging based on nearest neighbor filtering

Abstract

Computational ghost imaging is a new optical imaging technique that obtains a reconstructed image of a target from a set of illumination patterns, but the quality of the reconstructed image is affected by ambient light noise. In this paper, two nearest neighbor filtered illumination patterns generated based on nonlinear noise filtering algorithm are proposed, which can effectively reduce the effect of ambient light in ghost imaging. When additional noise is added, compared with the reconstructed images of random illumination pattern, Hadamard illumination pattern and linear filter illumination pattern, the SNR of the reconstructed images of two different neighbor filtering illumination patterns is increased by 470 %, 1863 %, 75 % and 517 %, 2022 %, 90 %, respectively. In simulated noise environments, neighbor filtering illumination patterns reduce most of the effects of noise and results in high quality reconstructed images for a variety of target objects. In the experiments, these new illumination patterns has a strong robustness, and the reconstructed images obtained are sharper and contain more detailed information than random illumination pattern, Hadamard illumination pattern, and linearly filtered illumination pattern. Nearest neighbor filtering illumination patterns apply the nonlinear filtering method to computational ghost imaging and obtains excellent imaging results, which provides a new denoising idea for computational ghost imaging.