Laser & Optoelectronics Progress, Volume. 61, Issue 22, 2211005(2024)

Polarization Image Reconstruction Method Based on Multi-View Camera Array

Yiming Ma1,2、*, Xin Wang1,2, XianKun Pu1,2, Lei Shi1,2, Han Han1,2, Yao Zha1,2, and Jun Gao1,2
Author Affiliations
  • 1School of Computer Science and Information Engineering, Hefei University of Technology, Hefei 230031, Anhui , China
  • 2Laboratory of Image Information Processing, Hefei University of Technology, Hefei 230031, Anhui , China
  • show less
    Figures & Tables(10)
    Polar-ReOccNet architecture diagram
    Multi-view polarization dataset scene construction and collection process.(a) Multi-view polarization camera array system; (b) occlusion scenario diagram; (c) schematic diagram of outdoor field data collection process
    Scene 1 (Fence 8 m). (a) Image after occlusion removal; (b) polarization degree image after occlusion removal; (c) polarization angle image after occlusion removal; (d) polarization degree and angle fusion image after occlusion removal; (e) original occluded image (f) original occluded polarization degree image; (g) original occluded polarization angle image; (h) original occluded polarization degree and angle fusion image; (i) ground truth image (j) ground truth polarization degree image; (k) ground truth polarization angle image; (l) ground truth polarization degree and angle fusion image
    Scene 2 (Fence 22 m). (a) Image after occlusion removal; (b) polarization degree image after occlusion removal; (c) polarization angle image after occlusion removal; (d) polarization degree and angle fusion image after occlusion removal; (e) original occluded image; (f) original occluded polarization degree image; (g) original occluded polarization angle image; (h) original occluded polarization degree and angle fusion image; (i) ground truth image; (j) ground truth polarization degree image; (k) ground truth polarization angle image; (l) ground truth polarization degree and angle fusion image
    Scene 3 (Bamboo 25 m). (a) Image after occlusion removal; (b) polarization degree image after occlusion removal; (c) polarization angle image after occlusion removal; (d) polarization degree and angle fusion image after occlusion removal; (e) original occluded image; (f) original occluded polarization degree image; (g) original occluded polarization angle image; (h) original occluded polarization degree and angle fusion image; (i) ground truth image; (j) ground truth polarization degree image; (k) ground truth polarization angle image; (l) ground truth polarization degree and angle fusion image
    • Table 1. Experimental scenes and scene parameters

      View table

      Table 1. Experimental scenes and scene parameters

      NumberData typeViewpoint countsDistance /mObject
      Ground truth5×155,6,7…50A&B
      1Fence_occ5×155,6,7…50A&B
      Fence5×155,6,7…50A&B
      2Banboo_occ5×155,6,7…50A&B
      Bamboo5×155,6,7…50A&B
      3Fence&bamboo_occ5×155,6,7…50A&B
      Fence&bamboo5×155,6,7…50A&B
      4Vine_occ5×155,5.5,6…24A&B
      Vine5×155,5.5,6…24A&B
      5Willow_occ5×155,5.5,6…24A&B
      Willow5×155,5.5,6…24A&B
    • Table 2. Value of the target after occlusion removal

      View table

      Table 2. Value of the target after occlusion removal

      NumberImageMSESSIMPSNR
      Scene 1Dop1.37610.651120.1006
      Aop5.66010.541314.2337
      DopAop1.31760.560520.6714
      Scene 2Dop1.64070.598319.4395
      Aop9.04410.468012.2418
      DopAop1.35670.579420.5443
      Scene 3Dop2.48390.539217.6283
      Aop1.26200.412310.7698
      DopAop1.98360.499218.8945
    • Table 3. Value of the target before occlusion removal

      View table

      Table 3. Value of the target before occlusion removal

      NumberImageMSESSIMPSNR
      Scene 1Dop2.45910.618717.6692
      Aop7.26590.527913.1490
      DopAop2.00100.590418.8568
      Scene 2Dop1.88310.641218.8408
      Aop17.44350.42789.3891
      DopAop2.21550.532818.4143
      Scene 3Dop2.70100.505717.2644
      Aop1.37490.412010.3977
      DopAop2.47320.496117.9365
    • Table 4. Features of GLCM of each target in each image after occlusion removal

      View table

      Table 4. Features of GLCM of each target in each image after occlusion removal

      GLCMContrastHomogeneityCorrelationEnergyEntropy
      Dop_A0.18840.94680.63570.75372.0731
      Aop_A1.55750.78560.88030.1554.0427
      DopAop_A0.22350. 91560.63390.50752.4936
      Dop_B0.97720.86310.67460.47333.0198
      Aop_B1.3670.82220.8890.22193.8278
      DopAop_B0.90680.86840.68730.36723.153
    • Table 5. Features of GLCM of each target in each image with original occlusion

      View table

      Table 5. Features of GLCM of each target in each image with original occlusion

      GLCMContrastHomogeneityCorrelationEnergyEntropy
      Dop_A_occ0.00440.99860.27740.99421.411
      Aop_A_occ2.88020.71090.7790.13354.328
      DopAop_A_occ0.00860.99620.37120.98181.4504
      Dop_B_occ0.02820.99360.25250.97511.4769
      Aop_B_occ1.75310.7960.87090.20823.8406
      DopAop_B_occ0.02810.98980.3840.95281.536
    Tools

    Get Citation

    Copy Citation Text

    Yiming Ma, Xin Wang, XianKun Pu, Lei Shi, Han Han, Yao Zha, Jun Gao. Polarization Image Reconstruction Method Based on Multi-View Camera Array[J]. Laser & Optoelectronics Progress, 2024, 61(22): 2211005

    Download Citation

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

    Category: Imaging Systems

    Received: Feb. 18, 2024

    Accepted: Apr. 12, 2024

    Published Online: Nov. 14, 2024

    The Author Email:

    DOI:10.3788/LOP240686

    CSTR:32186.14.LOP240686

    Topics