Fig. 1. (a) Schematic of our setup. The full-perpendicular polarized light is used for illumination, the object is hidden behind the scattering media, and the polarization camera is used for the sub-polarized images. (b) The intensity of the object and background as a function of different rotated angles of the polarizer. The intensity of polarized components is significantly larger than that of non-polarized components. (c₁), (c₃) Series of captured images with different rotated angle polarizers through the dynamic and static scattering media. The fresh chicken breast tissue and the ground glass with 600 grit are used for imaging. (c₂), (c₄) Normalized mean intensity values at the six different selected positions are a function of the angle of the polarizer. (d₁), (d₂) The distributions of DoLP and AoP. (d₃), (d₄) The statistical distributions of DoLP and AoP.

Fig. 2. (a) Flow chart of the proposed process. (a₁) Sample estimate. (a₂) The separated scattering media and target images. (a₃) The updating process. (a₄) Repeat for other measurements. (a₅) The high-resolution reconstructions. Imaging performance of the proposed platform. (b₁) Raw image (PSI) of USAF target through fresh eggshell membranes. (b₂) PDI results of USAF target through fresh eggshell membranes. (b₃) Iscat of the USAF target. (b₄) Isum of the USAF target. (b₅) The descattering results through orthogonal polarization characteristics. (b₆) Our reconstruction result. Imaging performance of the detailed information. (c₁) and (c₂) Intensity profiles across group 2, element 6 of images (b₁) and (b₆). (d₁) and (d₂) ESF of the selected region with a green rectangle. (e₁) and (e2) Fourier spectra of images (b₁) and (b₆).

Fig. 3. Underwater imaging results with 38, 80, and 154 NTU turbidity. (a) Schematic of underwater imaging setup; (b)–(d) Directly captured raw images, reconstructed images, zoom-ins of the ROIs ① and ②, and the DoP images in turbid water at 38, 80, and 154 NTU.

Fig. 4. Recovered results by the proposed platform for different objects. (a₁)–(b₁) and (c₁)–(d₁) The captured raw intensity images (paper and metallic material) through the dynamic and static scattering media, separately. (a₂)–(d₂) The recovered images using the proposed platform. (a₃)–(d₃) The detailed information in the first two columns.

Fig. 5. High-resolution reconstruction results by our method through chicken breast tissue as dynamic scattering media. (a) Raw intensity image. (b) Reconstruction results of four different time intervals (1, 2, 4, and 6 s) utilizing 50 pairs of sub-polarized images. (c) Reconstruction results of six different time intervals (1, 2, 4, 6, 10, and 30 s) utilizing 2, 5, and 10 pairs of sub-polarized images separately. (d)–(e) Intensity profiles across group 2, element 4 of six different time intervals (1, 2, 4, 6, 10, and 30 s) of five pairs, and four pairs (2, 5, 10, and 50) in 2 s time intervals, separately. (f)–(h) Evaluation parameter PSNR, SSIM, and mean square error (MSE) are used to describe the quality of the reconstructed images.

Fig. 6. High-resolution reconstruction results by our method through the ground glass as static scattering media. (a₁)–(c₁) Raw intensity images captured by the camera with 220, 600, and 1500 grit ground glass. (a₂)–(c₂) High-resolution reconstruction results of images (a₁)–(c₁). The quality of the reconstructed images is obviously improved. (a₃)–(c₃) The details of spectrum information distribution are in the top right corner of the first two columns. (d₁) and (d₂) Intensity profiles across the group 0, element 5 and group 1, element 5 of images (a₂)–(c₂).

Fig. 7. Influence of the number of input images on reconstruction quality. (a₁)–(a₆) Reconstruction results with different numbers of images. (b₁)–(b₆) Magnified details of local regions. (c) Stripe contrast of group 2, element 1.