Laser & Optoelectronics Progress, Volume. 62, Issue 14, 1439002(2025)
Multiscale Adversarial-Based Reconstruction Method for Occluded Polarized Images
Figures & Tables(10)
Fig. 2. Self-developed multi-view polarization array image acquisition system. (a) Polarimetric camera array system; (b) schematic diagram of field data collection process; (c) schematic diagram of occlusion scene with foreground occlusion and occluded targets; (d) overall schematic diagram of field data collection scene; (e) overall schematic diagram of indoor data collection scene
Fig. 3. Reconstruction results under outdoor fence occlusion for targets at different positions
Fig. 4. Reconstruction results under indoor bamboo occlusion for target at distance of 6 m
Table 1. Experimental scene parameters
View tableTable 1. Experimental scene parameters
Number Data type View point Distance /m Target 1 Ground truth 5×15 6‒70 A， B 2 Fence_com 5×15 6‒70 A， B 3 Bamboo_com 5×15 6‒70 A， B 4 Fence & Bamboo_com 5×15 6‒70 A， B 5 Vine_com 5×15 5‒24 A， B 6 Willow_com 5×15 5‒24 A， B
Table 2. Quantitative results of reconstructed images for different polarization directions, DoP, AoP, and fused polarization state
View tableTable 2. Quantitative results of reconstructed images for different polarization directions, DoP, AoP, and fused polarization state
Image Indicator Bamboo_com Fence_com Willow_com Restoration PSNR /dB 20.8437 25.9075 18.4094 SSIM 0.7201 0.7574 0.6503 MSE /10-8 3.1314 3.8603 1.3076 DoP PSNR /dB 20.6875 26.1862 24.0163 SSIM 0.7188 0.7890 0.6841 MSE /10-8 1.3127 3.7009 6.0994 AoP PSNR /dB 14.5862 18.0198 13.8636 SSIM 0.5617 0.6220 0.5255 MSE /10-8 4.9604 2.2499 6.7227 Fusion PSNR /dB 28.6171 28.2673 26.9208 SSIM 0.8005 0.7997 0.7207 MSE /10-8 2.1145 2.2919 3.1249
Table 3. GLCM feature values
View tableTable 3. GLCM feature values
Image Type Contrast Homogeneity Correlation Energy Entropy Occluded DoP 0.5637 0.9391 0.9731 0.3920 2.7798 AoP 0.8186 0.8927 0.7264 0.2361 3.3281 Fusion 0.5178 0.9508 0.9764 0.4136 2.6407 Restoration DoP 0.5122 0.9513 0.9742 0.3864 2.7910 AoP 0.5588 0.9343 0.7968 0.2638 3.1330 Fusion 0.4823 0.9642 0.9778 0.4125 2.5934 Ground truth DoP 0.5648 0.9370 0.9726 0.3867 2.8007 AoP 0.7096 0.8999 0.7161 0.2471 3.2126 Fusion 0.5435 0.9409 0.9751 0.4060 2.6768
Table 4. Quantitative comparison of reconstruction results for different algorithms
View tableTable 4. Quantitative comparison of reconstruction results for different algorithms
Algorithm PSNR /dB SSIM 15 m 40 m 65 m 15 m 40 m 65 m DeOccNet 14.6699 13.6252 0.4975 0.2401 0.3012 0.0484 MDFNet 14.4537 13.5857 0.6659 0.2315 0.2881 0.0378 Polar-ReOccNet 14.6418 13.6673 0.8719 0.2637 0.3090 0.0655 Proposed 20.8437 25.9075 18.4095 0.6928 0.7573 0.6503
Table 5. Quantitative results of ablation study
View tableTable 5. Quantitative results of ablation study
Algorithm PSNR /dB SSIM Bamboo_com Fence_com Willow_com Average Bamboo_com Fence_com Willow_com Average No-FEM 21.8543 23.0187 22.8647 22.5792 0.6219 0.6764 0.6170 0.6384 No-MHSA 22.4399 24.6695 22.9281 23.3458 0.6848 0.7194 0.6476 0.6839 L1 23.3828 25.5476 24.1916 24.3740 0.7467 0.7821 0.7251 0.7513 Proposed 24.2420 26.7870 24.7191 25.2494 0.7782 0.7962 0.7415 0.7720
Tools
Get Citation
Copy Citation Text
Han Han, Xin Wang, Xiankun Pu, Peifeng Pan, Yao Zha, Yajun Xu, Jun Gao. Multiscale Adversarial-Based Reconstruction Method for Occluded Polarized Images[J]. Laser & Optoelectronics Progress, 2025, 62(14): 1439002
Paper Information
Category: AI for Optics
Received: Jan. 16, 2025
Accepted: Mar. 2, 2025
Published Online: Jul. 2, 2025
The Author Email: Xin Wang (wangxin@.hfut.edu.cn)
CSTR:32186.14.LOP250525
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20