Pseudo-thermal ghost imaging with "learned" wavelength conversion

Abstract

Ghost imaging (GI) is an imaging modality using light that has never physically interacted with the object to be imaged. The success of GI relies on the strong spatial correlation of photons. However, not all optical systems in nature are strongly spatially correlated. Two-color pseudo-thermal GI (PGI) can be viewed as such a weakly correlated system with two independent light sources. In this Letter, Deep Learning is introduced to learn the correlation between two-color speckle patterns, which solves the problem of two-color PGI with a wavelength gap of 101 nm (from 633 nm to 532 nm). Further, we retrieved dual-band ghost images using one broad-spectrum bucket detector and the reference speckle patterns at 633 nm. Our scheme provides insights into all PGI with weak correlation and also is a potential approach for multi-spectral PGI with "learned"wavelength conversion, especially for invisible wavebands.