Optics and Precision Engineering, Volume. 29, Issue 11, 2692(2021)

Dehazing using a decomposition-composition and recurrent refinement network based on the physical imaging model

Yan-ru FENG1 and Yi-bin WANG2、*
Author Affiliations
  • 1School of Information Engineering, Institute of Disaster Prevention, Sanhe06520, China
  • 2School of Engineering, Sichuan Normal University, Chengdu610068, China
  • show less
    Figures & Tables(11)
    Flowchart of decomposition-composition and recurrent refinement network based on the physical imaging model
    The architecture of decomposition-composition and recurrent refinement network based on physical imaging model
    The architecture of LSTM recurrent unit
    The work mechanism of LSTM units
    The comparison of dehazing results on synthetic images(The PSNR/SSIM values are marked under each image with best results in bold)
    The comparison of dehazing results on synthetic images
    • Table 1. Different parameters setting and network architecture in terms of SSIM and PSNR

      View table
      View in Article

      Table 1. Different parameters setting and network architecture in terms of SSIM and PSNR

      n=1n=2n=3n=4n=5w/oHtw/op
      PSNR29.9830.5531.2231.2331.2429.7830.59
      SSIM0.9600.9680.9760.9760.9760.9580.974
    • Table 2. Transmission map by different algorithms in terms of SSIM

      View table
      View in Article

      Table 2. Transmission map by different algorithms in terms of SSIM

      DCP2BCCR4WAP5DCPDN6Ours
      Test O0.8660.8410.8860.9610.972
      SOTS0.8440.8300.8670.9560.973
    • Table 3. Dehazing results by different algorithms

      View table
      View in Article

      Table 3. Dehazing results by different algorithms

      DADN9DCPDN6GCAN8DNPAB10RDPN7Ours
      Test OPSNR25.0428.6724.1028.8128.9430.23
      SSIM0.930 70.942 80.929 60.961 00.974 90.975 9
      SOTSPSNR27.7617.8926.2827.3623.2431.22
      SSIM0.930 00.831 20.944 60.949 00.928 50.976 1
    • Table 4. Dehazing results by different algorithms

      View table
      View in Article

      Table 4. Dehazing results by different algorithms

      DADN9DCPDN6GCAN8DNPAB10RDPN7Ours
      CG0.2890.3470.2960.2740.2540.412
      VCM46.19555.19649.68047.98844.52660.114
      ENT15.01516.02415.89615.67214.89616.515
    • Table 5. Average running time of different methods processing single image

      View table
      View in Article

      Table 5. Average running time of different methods processing single image

      MethodsDADN9DCPDN6GCAN8DCPAB10DCP2Ours
      Time/s0.1670.0560.0090.0481.4260.037
    Tools

    Get Citation

    Copy Citation Text

    Yan-ru FENG, Yi-bin WANG. Dehazing using a decomposition-composition and recurrent refinement network based on the physical imaging model[J]. Optics and Precision Engineering, 2021, 29(11): 2692

    Download Citation

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

    Category: Information Sciences

    Received: May. 21, 2021

    Accepted: --

    Published Online: Dec. 10, 2021

    The Author Email: Yi-bin WANG (yibeen.wong@gmail.com)

    DOI:10.37188/OPE.20212911.2692

    Topics