Infrared and Laser Engineering, Volume. 51, Issue 11, 20220060(2022)

Multi-drop attention residual infrared image denoising network based on guided filtering

Jun Zhang1,2,3, Biao Zhu1,2,3, Yuzhen Shen1,2,3, and Peng Zhang1,2,3
Author Affiliations
  • 1Aviation Industry Corporation Huadong Photoelectric Company Limited, Wuhu 241002, China
  • 2State Special Display Engineering Laboratory, Wuhu 241002, China
  • 3National Special Display Engineering Research Center, Wuhu 241002, China
  • show less
    Figures & Tables(11)
    Diagram of spatial attention structure
    Diagram of channel attention structure
    Diagram of multi-drop attention resnet structure
    Diagram of GFC block structure
    Diagram of GFDNet structure
    Iraytek infrared noise data ((a), (c), (e), (g) are infrared noise images, (b), (d), (f), (h) are infrared clean images)
    Experimental results of the Iraytek dataset, the green area in the lower right corner is the magnification of the red area
    Experimental results of OTCBVS dataset, the green area in the lower right corner is the magnification of the red area
    • Table 1. Comparison table of loss weight

      View table
      View in Article

      Table 1. Comparison table of loss weight

      Loss weight200 iteration400 iteration
      PSRN/dBSSIMPSRN/dBSSIM
      0.115.290.372921.690.6729
      0.1215.340.384121.830.6841
      0.1415.390.400522.240.7013
      0.1615.410.399122.290.7004
      0.215.480.373922.010.6954
      0.2515.360.371422.110.6932
      0.515.170.369221.780.6689
    • Table 2. Comparison table of effective denoising experimental data

      View table
      View in Article

      Table 2. Comparison table of effective denoising experimental data

      AlgorithmYearIraytek dataset(300)OTCBVS(100)
      PSRN/dBSSIMPSRN/dBSSIM
      BM3D200727.430.7814__
      CBDNet201929.860.851727.870.7765
      RIDNet201930.670.899332.730.7821
      DPIR202030.780.900932.860.7839
      MIRNet202031.520.903232.770.7842
      Ours30.710.905732.710.7827
    • Table 3. Comparison table of ablation experiment indexes under Iraytek testset

      View table
      View in Article

      Table 3. Comparison table of ablation experiment indexes under Iraytek testset

      Model structureIraytek testset
      PSRN/dBSSIM
      Model 1.(UNet + conv + concat)27.960.8253
      Model 2.(UNet + conv + GFC)28.140.8432
      Model 3.(UNet+bottleneck+concat)28.860.8628
      Model 4.(UNet+ bottleneck + GFC)29.690.8861
      Model 5.(UNet +MAR + GFC)30.710.9057
    Tools

    Get Citation

    Copy Citation Text

    Jun Zhang, Biao Zhu, Yuzhen Shen, Peng Zhang. Multi-drop attention residual infrared image denoising network based on guided filtering[J]. Infrared and Laser Engineering, 2022, 51(11): 20220060

    Download Citation

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

    Category: Infrared technology and application

    Received: Jan. 18, 2022

    Accepted: --

    Published Online: Feb. 9, 2023

    The Author Email:

    DOI:10.3788/IRLA20220060

    Topics