Optics and Precision Engineering, Volume. 30, Issue 19, 2404(2022)

Polarization computational imaging super-resolution reconstruction with lightweight attention cascading network

Jie WANG1... Guoming XU1,2,3,*, Jian MA1,2, Yong WANG3 and Yi LI4 |Show fewer author(s)
Author Affiliations
  • 1School of Internet, Anhui University, Hefei230039, China
  • 2National Engineering Research Center for Agro-Ecological Big Data Analysis & Application, Anhui University, Hefei30601, China
  • 3Anhui Province Key Laboratory of Polarized Imaging Detecting Technology, Army Artillery and Air Defense Forces Academy of PLA, Hefei2001, China
  • 4Institute of Intelligent Technology, Anhui Wenda University of Information Engineering, Hefei231201, China
  • show less
    Figures & Tables(20)
    Network architecture of lightweight attention cascading network
    Network structure of spatial pyramid
    Network structure of enhanced spatial attention block
    Network structure of channel attention mechanism
    Network structure of cascading attention block
    Spectral polarization camera
    Polarization images of building
    Polarization images of hefei south railway station
    Different polarization images
    Comparisons of the accuracy and model parameters
    Comparisons of the rebuild performance and speed
    Visualized results of ×2 SR on monument fully polarization image
    Visualized results of ×3 SR on airport runway fully polarization image
    Visualized results of ×4 SR on building fully polarization image
    • Table 1. 空间注意力网络对重建结果的影响(4 SR)

      View table
      View in Article

      Table 1. 空间注意力网络对重建结果的影响(4 SR)

      ESAPSNR/dBSSIM
      ×38.860.944 3
      38.930.944 7
    • Table 2. MSAB模块对重建效果的影响(4 SR)

      View table
      View in Article

      Table 2. MSAB模块对重建效果的影响(4 SR)

      MSABPSNR/dBSSIM
      ×38.870.944 1
      38.930.944 7
    • Table 3. 信息细化块对重建性能的影响(4 SR)

      View table
      View in Article

      Table 3. 信息细化块对重建性能的影响(4 SR)

      信息细化块PSNR/dBSSIM
      ×38.810.943 9
      38.930.944 7
    • Table 4. 不同路径的重建模块客观效果对比(4 SR)

      View table
      View in Article

      Table 4. 不同路径的重建模块客观效果对比(4 SR)

      重建模块路径PSNR/dBSSIM
      138.850.944 3
      238.930.944 7
    • Table 5. 不同网络深度对重建效果的影响(3 SR)

      View table
      View in Article

      Table 5. 不同网络深度对重建效果的影响(3 SR)

      CAB数量Parameters/KPSNR/dBSSIM
      1048441.800.960 6
      1151941.760.960 4
      1255441.820.960 6
      1358941.790.960 7
      1462441.810.960 6
      1565941.830.960 8
    • Table 6. Indicator comparison of different SR algorithms on fully polarization image set

      View table
      View in Article

      Table 6. Indicator comparison of different SR algorithms on fully polarization image set

      ScaleMethod

      Parameters

      /K

      Monument imageBridge imageBuilding imageRoad image
      PSNR/dBSSIMPSNR/dBSSIMPSNR/dBSSIMPSNR/dBSSIM
      ×2SRCNN5744.920.988 543.820.972 743.380.974 942.920.965 7
      FSRCNN1244.770.988 043.850.972 943.390.974 942.880.965 4
      MSRN593046.740.989 645.120.977 544.630.978 744.110.971 1
      AWSRN139646.740.989 645.120.977 644.640.978 744.120.971 2
      Ours54346.740.989 645.120.977 644.630.978 744.120.971 2
      ×3SRCNN5741.140.968 441.550.957 640.970.958 340.990.946 4
      FSRCNN1241.130.968 341.820.958 941.130.959 138.680.948 0
      MSRN611443.260.975 943.370.967 942.770.967 842.390.957 1
      AWSRN147643.310.976 143.380.968 142.760.967 842.390.957 2
      Ours55443.250.975 943.390.968 142.780.967 942.390.957 2
      ×4SRCNN5737.640.939 338.450.932 838.050.930 338.890.919 4
      FSRCNN1237.690.938 539.360.940 138.420.932 838.960.919 3
      MSRN607840.260.957 741.410.957 140.400.950 940.680.939 7
      AWSRN158740.300.958 041.900.959 840.340.950 640.670.939 5
      Ours57640.210.957 441.920.959 640.580.952 240.650.939 3
    Tools

    Get Citation

    Copy Citation Text

    Jie WANG, Guoming XU, Jian MA, Yong WANG, Yi LI. Polarization computational imaging super-resolution reconstruction with lightweight attention cascading network[J]. Optics and Precision Engineering, 2022, 30(19): 2404

    Download Citation

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

    Category: Information Sciences

    Received: May. 13, 2022

    Accepted: --

    Published Online: Oct. 27, 2022

    The Author Email: XU Guoming (xgm121@163.com)

    DOI:10.37188/OPE.20223019.2404

    Topics