Fig. 1. Common modulators in ghost imaging

Fig. 2. Structural diagram of ghost imaging based on projector. (a) Ghost imaging and (b) single-pixel imaging when object to be measured is transmissive and modulator is transmissive; (c) ghost imaging and (d) single-pixel imaging when object to be measured is transmissive and modulator is reflective; (e) ghost imaging and (f) single-pixel imaging when object to be measured is reflective and modulator is transmissive; (g) ghost imaging and (h) single-pixel imaging when object to be measured is reflective and modulator is reflective

Fig. 3. Structural diagram of ghost imaging based on projector

Fig. 4. Key technology of ghost imaging

Fig. 5. Experimental setup and simulation results of underwater ghost imaging. (a) Experimental setup of underwater ghost imaging; (b) ghost imaging, differential ghost imaging, compressed sensing ghost imaging and convolutional neural network ghost imaging (CNNGI) in water environment added salt and milk

Fig. 6. Underwater ghost imaging results under backscatter light^[52]. (a) Simulation result; (b) experiment result

Fig. 7. Effects of underwater speckle-object distance, water turbidity and angle of view on underwater ghost imaging^[9]. (a) Experimental setup of underwater imaging; (b) experimental setup for measurement of angle of view; (c) images with different turbidities captured by charge-coupled device (CCD); (d) ghost imaging results at different frame numbers, turbidities, and positions; (e) relationship between signal-to-noise ratio and turbidity;(f) relationship among optical power, signal-to-noise ratio and angle of view

Fig. 8. Effects of speckle-object distance and refractive index on underwater ghost imaging^[10]. (a) Ghost imaging results at different distances; (b) width of point spread function varying with total optical path under different distances; (c) ghost imaging results under different refractive indices; (d) width of point spread function varying with total optical path under different refractive indices

Fig. 9. Effects of underwater speckle-object distance and water temperature on underwater ghost imaging^[53]. (a) Light path diagram; (b) imaging results under different locations and temperatures; (c) relationship between signal-to-noise ratio of ghost imaging and temperature at different locations; (d) relationship between signal-to-noise ratio of ghost imaging and speckle-object distance

Fig. 10. Effects of temperature gradient and vibration amplitude on underwater ghost imaging^[11]. (a) Schematic diagram of underwater environment; (b) imaging results under different locations and temperature gradients; (c) relationship between structural similarity (SSIM) and temperature gradient at different locations; (d) imaging results and structural similarities under different vibration amplitudes; (e) relationship between structural similarity and turbidity

Fig. 11. Effects of propagation distance, ocean turbulence coefficients and beam morphology on underwater ghost imaging^{[55, 57]}. (a) Imaging results under different propagation distances; (b) root mean square error varying with ε,χT and ω; (c) ghost imaging results of Gaussian light source and Lorentz light source; (d) relationship between root mean square error and ocean turbulence parameters under Gaussian light source and Lorentz light source with different modulation factors

Fig. 12. Effects of ocean turbulence coefficient and turbidity on underwater ghost imaging^[58-59]. (a) Relationship between visibility and χT; (b) relationship between visibility and ε; (c) relationship between visibility and ω; (d) relationship between visibility and wavelength; (e) relationship between visibility and coherence length; (f) traditional imaging and ghost imaging under different turbidities; (g) relationship between peak signal-to-noise ratio and relative attenuation coefficient; (h) relationship between visibility and relative attenuation coefficient

Fig. 13. Effects of incident angle, propagation distance and ocean turbulence coefficients on underwater ghost imaging^[60]. (a) Light path diagram of underwater; (b) imaging results of different incident angles; (c) imaging results of different transmission distances; (d) relationship between visibility and χT; (e) relationship between visibility and ε; (f) relationship between visibility and ω

Fig. 14. Hadamard speckle underwater ghost imaging^[61]. (a) Imaging results under random speckles and Hadamard speckles; (b) comparison between Hadamard speckle ghost imaging and traditional imaging under different turbidities

Fig. 15. Effects of laser power, projection rate and turbidity on Hadamard underwater ghost imaging^[62]. (a) Hadamard underwater ghost imaging under different laser powers; (b) Hadamard underwater ghost imaging under different projection frequencies; (c) Hadamard underwater ghost imaging under different turbidities; (d) comparison of Hadamard underwater ghost imaging, camera, compressed sensing ghost imaging and Fourier ghost imaging results

Fig. 16. Pulse peak detection underwater ghost imaging^[63]. (a) Ghost imaging at different transmission distances; (b) ghost imaging at different incident angles; (c) ghost imaging at different medium attenuation coefficients

Fig. 17. Push-sweep computing underwater ghost imaging and cross-polarization underwater ghost imaging^[64-66]. (a) Structural diagram of push-sweep computing underwater ghost imaging; (b) imaging comparison under different turbidities and transmission distances; (c) structural diagram of cross-polarization underwater ghost imaging; (d) imaging comparison under different polarizations and turbidities

Fig. 18. Underwater polarization differential ghost imaging based on histogram preprocessing^[12]. (a) Schematic diagram of light path of histogram polarization differential ghost imaging; (b) ghost imaging results under different turbidities and polarizations; (c) histogram statistical results of gray values of reconstructed images under different concentrations

Fig. 19. Underwater ghost imaging for convex set alternating projection^[8]

Fig. 20. Underwater ghost imaging compensated by water degradation function^[67]. (a) Comparison of ghost imaging compensated by water degradation function, differential ghost imaging and Fourier ghost imaging; (b) results of ghost imaging compensated by water degradation function and Fourier ghost imaging at different sampling rates

Fig. 21. Underwater compression computing ghost imaging based on wavelet enhancement^[68]. (a) Imaging results of method in Ref.[68], "Russian Doll" method, method based on four connected region, total variation method, low grade method, and "Cake Cutting" method in tap water; (b) imaging results of method in Ref.[68], "Russian Doll" method, method based on four connected region, total variation method, low grade method, and "Cake Cutting" method in turbid water; (c) relationship between peak signal-to-noise ratio and sampling rate in tap water; (d) relationship between peak signal-to-noise ratio and sampling rate in turbid water

Fig. 22. Underwater ghost imaging based on compressed sensing super-resolution convolutional neural network^[69]. (a) Schematic diagram of reconstruction based on compressed sensing super-resolution convolutional neural network; (b) simulation results under different sampling rates; (c) simulation comparison of method in Ref.[69], ghost imaging and super-resolution convolutional neural network under different turbidities; (d) experimental result comparison of method in Ref. [69], ghost imaging and super-resolution convolutional neural network under different turbidities; (e) comparison between method in Ref. [69] and total variational regularization under different sampling rates

Fig. 23. Underwater ghost imaging based on generative adversarial network^[70]. (a) Experimental setup of underwater ghost imaging based on generative adversarial network; (b) Cycle-GAN network model structure for generating paired underwater data sets; (c) paired underwater data sets;(d) model structure of generative adversarial network; (e) simulation comparison of underwater ghost imaging based on generative adversarial network, pyramid deep learning ghost imaging and U-NET deep learning ghost imaging; (f) experimental comparison of underwater ghost imaging based on generative adversarial network, pyramid deep learning ghost imaging and U-NET deep learning ghost imaging; (g) influence of turbidity on underwater ghost imaging based on generative adversarial network, pyramid deep learning ghost imaging and U-NET deep learning ghost imaging