Ghost imaging systems use down-conversion sources that produce twin output beams of position-correlated photons to produce an image of an object using photons that did not interact with the object. One of these beams illuminates the object and is detected by a single pixel detector while the image information is recovered from the second, spatially correlated, beam. We utilize this technique to obtain images of objects probed with 1.5μm photons whilst developing the image using a highly efficient, low-noise, photon-counting camera detecting the correlated photons at 460nm. The efficient transfer of the image information from infrared illumination to visible detection wavelengths and the ability to count single-photons allows the acquisition of an image while illuminating the object with an optical power density of only 100 pJ cm<sup>-2</sup> s<sup>-1</sup>. We apply image reconstruction techniques based on compressive sensing to reconstruct our images from data sets containing far fewer photons than conventionally required. This wavelength-transforming ghost imaging technique has potential for the imaging of light-sensitive specimens or where covert operation is desired.
How many photons does it take to form an image? Although a single photon can be spatially encoded to carry large amounts of information, real images are not fully orthogonal to each other and hence, realistically, require many detected photons to distinguish between them. Even if one has access to a pixelated imaging detector with high quantum efficiency, the fidelity of a recorded, or inferred, image depends critically upon the dark counts from the detector. Here we present imaging using heralded single-photons and a time-gated intensified camera to all but eliminate noise-events, and record images of a standard test-target. The images are formed from only a few thousand photons and are therefore subject to a noise inherent within the Poissonian distribution of single-photon events. We apply techniques of compressive sensing and image regularization to obtain good estimates of the object, obtained for ultra-low optical exposures.