Fig. 1. Schematic diagram of the MPFNet.

Fig. 2. Schematic diagram of the MBF module.

Fig. 3. Schematic diagram of the MBSCA structure.

Fig. 4. Schematic diagram of the experimental setup. (a) The free-space ghost imaging experimental setup. The laser passes through a lens and shines on the DMD. The laser passes through the object and is received by the SPD. (b) The underwater ghost imaging experimental setup. The light source is modulated by a PBS and a QWP. The dashed box in the first part represents the acquisition of linear polarization signals, and the dashed box in the back represents the acquisition of circular polarization signals.

Fig. 5. Reconstruction results in free space, with the corresponding grayscale and ground truth of the object.

Fig. 6. Quantitative evaluations of the CNR, PSNR, and resolution of TGI, DGI, and the MPFNet. Object-DGI and Object-TGI represent the reconstruction results of DGI and TGI at different sampling rates. Object-Ours-s represents the single input and Object-Ours-m represents the dual-branch input.

Fig. 7. (a) Relationship between CNR and resolution for images reconstructed by the network using different speckle sizes. (b) Relationship between CNR and resolution for DGI reconstruction results of light fields modulated with various speckle sizes in underwater environments.

Fig. 8. Reconstructed image results using simulated data. Ours-s represents single-branch imaging results; Ours-m represents dual-branch imaging results including MBF and MBSCA structures.

Fig. 9. (a) Reconstructed underwater ghost images and corresponding CNR, PSNR, and resolution. ResUNet-LP-∥, ResUNet-LP-⊥, and ResUNet-CP represent the results of using ResUNet to image the two vertical components of linear polarization and circular polarization data, respectively. GIDC-LP-∥, GIDC-LP-⊥, and GIDC-CP represent the results of imaging the two perpendicular components of linear polarization and circular polarization data using GIDC, respectively. LP-∥ represents the result with horizontal linear polarization. LP-⊥ represents the result with vertical linear polarization. Multi-LP represents the fusion of perpendicular linear polarization components. CP represents circular polarization. LP + CP represents the result of the fusion of linear and circular polarizations. We highlight the imaging results of ResUNet and GIDC and the imaging results of our proposed method with the green solid box and the red solid box. (b) The fluctuations of 1500 data points collected for the letter "P" under two linear polarization directions.