Fig. 1. Flow chart of image classification-restoration method used for computational ghost imaging

Fig. 2. Classification network structure

Fig. 3. Network structure of generator

Fig. 4. Network structure of discriminator

Fig. 5. Schematic diagram of computational ghost imaging through atmospheric turbulence

Fig. 6. Influence of atmospheric turbulence of different intensities on computational ghost imaging. (a) Initial speckle field and object; (b) speckle field and reconstructed images without turbulence; (c)--(g) speckle field and reconstructed images with different turbulence strengths

Fig. 7. Classified results of blurred images by classification network

Fig. 8. Simulation results of restoration network. (a) Blurred images; (b) restored images using classification-restoration network; (c) restored images only using restoration network

Fig. 9. Image evaluation index of blurred images and restored images under each category. (a) PSNR mean value; (b) SSIM mean value

Fig. 10. Restoration results for other types of images