Ghost imaging (GI), which is a statistical imaging method from quantum optics, is positioned as a kind
of single pixel imaging, where imaging is performed at a single pixel. Due to its statistical processing, it
has excellent properties for imaging in disturbed environments and with low light levels. Unfortunately,
it suffers from long measurement times caused by the necessity of multiple illuminations. Because of
these problems, many reports have focused on compressive sensing that analytically performs imaging.
Recently, the introduction of machine learning is fueling attention on GI. In this paper, we comprehensively
explain GI from the origin of quantum optics to its transition and classification with the introduction
of machine learning.
View full abstract