Acta Photonica Sinica, Volume. 52, Issue 11, 1110003(2023)

Infrared and Visible Image Fusion Based on Dual Channel Residual Dense Network

Xin FENG, Jieming YANG*, Hongde ZHANG, and Guohang QIU
Author Affiliations
  • School of Mechanical Engineering,Key Laboratory of Manufacturing Equipment Mechanism Design and Controlof Chongqing,Chongqing Technology and Business University,Chongqing 400067,China
  • show less
    Figures & Tables(11)
    Auto-Encoder structure framework
    Encoder network structure
    A Residual dense network fusion framework based on dual-channel cascade
    Infrared and visible light fusion results of different channels
    Infrared and visible light fusion results of different fusion strategies
    Fusion results of“pedestrian”
    Infrared and visible light fusion results of each algorithm
    Objective experimental line chart
    • Table 1. The table of network structure

      View table
      View in Article

      Table 1. The table of network structure

      LayerSizeStrideInputOutputActivation
      EncoderConv31116ELU
      Dual path block×3-----
      DecoderResConv313216ELU
      DenConv316464-
      Conv316432-
      Conv313216ELU
      GeneralConv313216-
      Conv31161Tanh
      Dual path blockConv11Res:1632ELU
      Den:1632
      Conv31Res:3232ELU
      Den:3232
      Conv11Res:3232ELU
      Den:3232
    • Table 2. The table of mean quantitative values of experimental evaluation indexes for ablation

      View table
      View in Article

      Table 2. The table of mean quantitative values of experimental evaluation indexes for ablation

      MetricsDifferent channelsDifferent fusion strategies
      ResDenseDual PathAddAvgMaxDual path attention
      SSIM0.906 00.912 30.923 60.912 30.905 90.891 80.928 9
      SCD1.712 91.702 81.718 91.696 51.555 41.535 01.860 5
      PSNR63.601 165.391 865.682 859.621 664.248 662.230 566.072 7
    • Table 3. The average quantitative value of each evaluation index

      View table
      View in Article

      Table 3. The average quantitative value of each evaluation index

      MetricsMethods
      MST-SRGTFIFCNNPMGIU2FusionDenseFuseRes2NetFusionGANOurs
      SD32.576 429.375 139.122 832.641 536.615 032.877 339.870 924.325 540.747 1
      CC0.577 80.437 10.585 30.561 60.598 30.610 20.558 60.493 50.604 5
      PSNR65.903 663.622 965.824 663.405 265.111 564.650 764.632 062.664 966.072 7
      MS-SSIM0.942 90.854 60.940 10.912 60.923 30.906 60.892 20.784 90.932 4
      SCD1.760 31.024 81.900 51.873 51.872 81.606 41.668 11.596 41.884 9
      MI1.879 32.398 92.413 62.139 32.178 62.438 72.823 22.070 43.123 8
    Tools

    Get Citation

    Copy Citation Text

    Xin FENG, Jieming YANG, Hongde ZHANG, Guohang QIU. Infrared and Visible Image Fusion Based on Dual Channel Residual Dense Network[J]. Acta Photonica Sinica, 2023, 52(11): 1110003

    Download Citation

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

    Category:

    Received: May. 19, 2023

    Accepted: Jun. 13, 2023

    Published Online: Dec. 22, 2023

    The Author Email: Jieming YANG (2871600119@qq.com)

    DOI:10.3788/gzxb20235211.1110003

    Topics