Laser & Optoelectronics Progress, Volume. 62, Issue 14, 1439002(2025)

Multiscale Adversarial-Based Reconstruction Method for Occluded Polarized Images

Han Han1,2, Xin Wang1,2、*, Xiankun Pu3, Peifeng Pan1,2, Yao Zha1,2, Yajun Xu1,2, and Jun Gao1,2
Author Affiliations
  • 1School of Computer Science and Information Engineering, Hefei University of Technology, Hefei 230009, Anhui , China
  • 2Laboratory of Image Information Processing, Hefei University of Technology, Hefei 230009, Anhui , China
  • 3School of Automotive and Traffic Engineering, Hubei University of Arts and Science, Xiangyang441053, Hubei , China
  • show less
    Figures & Tables(10)
    Structure of PolarReconGAN network and included modules
    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
    Reconstruction results under outdoor fence occlusion for targets at different positions
    Reconstruction results under indoor bamboo occlusion for target at distance of 6 m
    Comparison experimental results of different algorithms
    • Table 1. Experimental scene parameters

      View table

      Table 1. Experimental scene parameters

      NumberData typeView pointDistance /mTarget
      1Ground truth5×156‒70A, B
      2Fence_com5×156‒70A, B
      3Bamboo_com5×156‒70A, B
      4Fence & Bamboo_com5×156‒70A, B
      5Vine_com5×155‒24A, B
      6Willow_com5×155‒24A, B
    • Table 2. Quantitative results of reconstructed images for different polarization directions, DoP, AoP, and fused polarization state

      View table

      Table 2. Quantitative results of reconstructed images for different polarization directions, DoP, AoP, and fused polarization state

      ImageIndicatorBamboo_comFence_comWillow_com
      RestorationPSNR /dB20.843725.907518.4094
      SSIM0.72010.75740.6503
      MSE /10-83.13143.86031.3076
      DoPPSNR /dB20.687526.186224.0163
      SSIM0.71880.78900.6841
      MSE /10-81.31273.70096.0994
      AoPPSNR /dB14.586218.019813.8636
      SSIM0.56170.62200.5255
      MSE /10-84.96042.24996.7227
      FusionPSNR /dB28.617128.267326.9208
      SSIM0.80050.79970.7207
      MSE /10-82.11452.29193.1249
    • Table 3. GLCM feature values

      View table

      Table 3. GLCM feature values

      ImageTypeContrastHomogeneityCorrelationEnergyEntropy
      OccludedDoP0.56370.93910.97310.39202.7798
      AoP0.81860.89270.72640.23613.3281
      Fusion0.51780.95080.97640.41362.6407
      RestorationDoP0.51220.95130.97420.38642.7910
      AoP0.55880.93430.79680.26383.1330
      Fusion0.48230.96420.97780.41252.5934
      Ground truthDoP0.56480.93700.97260.38672.8007
      AoP0.70960.89990.71610.24713.2126
      Fusion0.54350.94090.97510.40602.6768
    • Table 4. Quantitative comparison of reconstruction results for different algorithms

      View table

      Table 4. Quantitative comparison of reconstruction results for different algorithms

      AlgorithmPSNR /dBSSIM
      15 m40 m65 m15 m40 m65 m
      DeOccNet14.669913.62520.49750.24010.30120.0484
      MDFNet14.453713.58570.66590.23150.28810.0378
      Polar-ReOccNet14.641813.66730.87190.26370.30900.0655
      Proposed20.843725.907518.40950.69280.75730.6503
    • Table 5. Quantitative results of ablation study

      View table

      Table 5. Quantitative results of ablation study

      AlgorithmPSNR /dBSSIM
      Bamboo_comFence_comWillow_comAverageBamboo_comFence_comWillow_comAverage
      No-FEM21.854323.018722.864722.57920.62190.67640.61700.6384
      No-MHSA22.439924.669522.928123.34580.68480.71940.64760.6839
      L123.382825.547624.191624.37400.74670.78210.72510.7513
      Proposed24.242026.787024.719125.24940.77820.79620.74150.7720
    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

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    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)

    DOI:10.3788/LOP250525

    CSTR:32186.14.LOP250525

    Topics