Fig. 1. Imaging process of ghost imaging

Fig. 2. Comparison of communication system and ghost imaging system

Fig. 3. Channel model for sampling process of ghost imaging

Fig. 4. R independent sub-channels after SVD

Fig. 5. Statistical characteristics analysis of imaging scene information. (a) Parts of imaging scene; (b) corresponding MSCN coefficient images; (c) fitting of MSCN coefficient distribution and Gaussian distribution of 100 imaging scene images (lines represent mean value, about 0.95); (d) parts of artificial image; (e) corresponding MSCN coefficient images; (f) MSCN coefficient distribution of 5 kinds of artificial images

Fig. 6. Variations of channel capacity and MSE with distribution type of matrix elements. (a) Imaging scene image used in simulation; (b) channel capacity; (c) MSE of reconstructed image under GPSR reconstruction algorithm; (d) MSE of reconstructed image under pseudo-inverse reconstruction algorithm

Fig. 7. Variations of channel capacity and MSE with number of samples. (a) Imaging scene images used in simulation; (b) channel capacity; (c) MSE of reconstructed image under GPSR reconstruction algorithm; (d) MSE of reconstructed image under pseudo-inverse reconstruction algorithm

Fig. 8. Comparison of normalized channel capacity and normalized inversion MSE under two reconstruction algorithms. (a) Three normalized curves; (b) GPSR reconstruction algorithm; (c) pseudo-inverse reconstruction algorithm

Fig. 9. Box-plots of all fits of 100 imaging scenes under two reconstruction algorithms. (a) GPSR reconstruction algorithm; (b) pseudo-inverse reconstruction algorithm