Fig. 1. Schematics of ultraviolet computational ghost imaging (UV-CGI). (a) Experimental setup. L1–L5, lenses; P, pinhole; DMD, digital micromirror device; M, mirror; ID, iris diaphragm; SPD, single-pixel detector. (b) UV-sensitive image of the object; the yellow part within the image represents the area that contains UV-sensitive samples. (c) As a comparison, traditional visible CGI cannot reveal such a UV image.

Fig. 2. Reconstructed images of three different samples for sampling ratios of 10%, 20%, 50%, and 100%. The left columns show the ground-truth photos of the samples: (a) sponge coated with sunscreen except in the central triangular area; (b) sponge coated with sunscreen except in an umbrella-shaped area with the letters “UV” written on it; (c) bottom part of a triangular piece of paper coated with sunscreen. (d) Schematic diagram of the spectral measurement. (e) Measurement spectra of the paper and the sponges under UV irradiation.

Fig. 3. Detection of sunscreen by UV-CGI. (a) Experimental procedure. (i) Sample preparation. (ii) UV-CGI exposures. (b) Visible light photos of the sample; the different colors indicate the four areas of the sample coated with different densities of sunscreen, areas 1–4 containing, respectively, 0, 0.2, 0.5, and 1 μL. (c) Reconstructed image of the sample, in which the four areas are shown with different grayscale values. (d) The smoothed result of (c). (e) Statistical grayscale values of the four areas in (d).

Fig. 4. Dark-field UV-CGI of transmissive pure phase objects (left side) and a comparison experiment in a bright-field (right side). (a) Experimental setup of transmissive dark-field UV-CGI. (b) Schematic of phase object and the recovered CGI image. (c)–(e) Reconstructed dark-field images of three samples for sampling ratios of 10%, 20%, 50%, and 100%. The left column shows the photos of two fingers holding the phase objects, which are the letters (c) “V”, (d) “L”, and (e) an exclamation mark “!”. (f) Experimental setup of bright-field UV-CGI. (g) Reconstructed bright-field images of three samples for 100% sampling ratio.

Fig. 5. Detection of the damages on a CD by dark-field UV-CGI (left side) and a comparative experiment in a bright-field UV-CGI (right side). (a) Experimental setup of reflective dark-field UV-CGI. (b) Photograph of the damaged CD. Three slightly damaged areas are shown in the colored boxes, corresponding to (1) a point-like scratch, (2) a line scratch, and (3) two line scratches. Another two heavily damaged areas are indicated by the black boxes as a comparison. (c)–(e) Reconstructed dark-field images of areas (1)–(3) in (b) for sampling ratios of 10%, 20%, 50%, and 100%. The left column shows the magnified photos of the three areas. (f) Experimental setup of bright-field UV-CGI. (g) Reconstructed bright-field images of the three areas for 100% sampling ratios.

Fig. 6. Dark-field UV-CGI with different kinds of beam stops in the transmission scheme. (a) Photo of the transparent square phase object, of size 3.5  mm×3.5  mm; its shape is shown in the upper right corner. The five different beam stops are shown in (b)–(f). (b) The beam stop has a size (about 3.7  mm×3.7  mm) approximately equal to the object’s size. (c) The beam stop has a size (about 8.0  mm×8.0  mm) that is larger than object’s size. (d) Same large beam stop used in (c) but moved such that the beam passes through its upper right corner. (e) Similar to (d), but with the beam passing through the lower left corner. (f) Circular beam stop of diameter 5.3 mm, as indicated by the yellow circular area in the inset; the square inside the yellow circle indicates the illumination beam, which encompasses the phase object. The reconstructed results corresponding to the cases (b)–(f) are shown in (g)–(k), with a sampling rate of 100%.

Fig. 7. Simulation results of dark-field UV-CGI with a pure phase object and comparison results in the bright-field condition. (a) Amplitude (upper) and phase (lower) of the original object. (b) PSNR and MSE curves of the reconstructed dark-field edge images versus sampling ratio, from 5% to 100%. (c) Example of reconstructed dark-field images with different sampling ratios. (d) Reconstructed bright-field images with the same sampling ratios as in (c).