Ghost imaging: from quantum to classical to computational

Abstract

Ghost-imaging experiments correlate the outputs from two photodetectors: a high-spatial-resolution (scanning pinhole or CCD array) detector that measures a field that has not interacted with the object to be imaged, and a bucket (single-pixel) detector that collects a field that has interacted with the object. We give a comprehensive review of ghost imaging—within a unified Gaussian-state framework—presenting detailed analyses of its resolution, field of view, image contrast, and signal-to-noise ratio behavior. We consider three classes of illumination: thermal-state (classical), biphoton-state (quantum), and classical-state phase-sensitive light. The first two have been employed in a variety of ghost-imaging demonstrations. The third is the classical Gaussian state that produces ghost images that most closely mimic those obtained from biphoton illumination. The insights we develop lead naturally to a new, single-beam approach to ghost imaging, called computational ghost imaging, in which only the bucket detector is required. We provide quantitative results while simultaneously emphasizing the underlying physics of ghost imaging. The key to developing the latter understanding lies in the coherence behavior of a pair of Gaussian-state light beams with either phase-insensitive or phase-sensitive cross correlation.