Laser & Optoelectronics Progress, Volume. 62, Issue 16, 1637005(2025)

Dynamic Contrast Dual-Branch Feature Decomposition Network for Infrared-Visible Image Fusion

Linglin Bao1, Pengge Ma1,2,3、*, Long Wang1, Jinwang Qian1, Zhaoyu Liu2,3, and Qiuchun Jin1
Author Affiliations
  • 1School of Electronics and Information, Zhengzhou University of Aeronautics, Zhengzhou 450046, Henan , China
  • 2Henan Province Key Laboratory of General Aviation Technology, Zhengzhou University of Aeronautics, Zhengzhou 450046, Henan , China
  • 3Henan Aerospace Electronic Information Technology Collaborative Innovation Center, Zhengzhou University of Aeronautics, Zhengzhou 450046, Henan , China
  • show less
    Figures & Tables(11)
    The framwork of the DCFN
    Codec branch training process
    Fusion layer training process
    Comparison of infrared-visible image fusion results on MSRS dataset. (a) Visible light image; (b) infrared image; (c) DIF fusion result; (d) ReCoNet fusion result; (e) RFNet fusion result; (f) IFNet fusion result; (g) FusionNet fusion result; (h) DeepFuse fusion result; (i) TFNet fusion result; (j) fusion result of proposed method
    Comparison of infrared-visible image fusion results on RoadScene dataset. (a) Visible light image; (b) infrared image; (c) DIF fusion result; (d) ReCoNet fusion result; (e) RFNet fusion result; (f) IFNet fusion result; (g) FusionNet fusion result; (h) DeepFuse fusion result; (i) TFNet fusion result; (j) fusion result of proposed method
    Comparison of infrared-visible image fusion results on TNO dataset. (a) Visible light image; (b) infrared image; (c) DIF fusion result; (d) ReCoNet fusion result; (e) RFNet fusion result; (f) IFNet fusion result; (g) FusionNet fusion result; (h) DeepFuse fusion result; (i) TFNet fusion result; (j) fusion result of proposed method
    • Table 1. MSRS dataset fusion results

      View table

      Table 1. MSRS dataset fusion results

      MethodENSDSFMIFMIVIFQAB/FSSIM
      DIF7.624061.31160.05481.21940.90270.01040.68490.5486
      ReCoNet7.457261.50090.06291.86340.87460.01070.66930.5627
      RFNet7.664562.07740.08701.35910.90300.01300.69110.5485
      IFNet7.676559.23060.06491.66840.82900.01180.64890.5855
      FusionNet4.864036.18500.01270.98641.30530.00290.61810.3473
      DeepFuse7.045138.90770.05291.42600.87360.01010.76150.5632
      TFNet7.511063.13000.06682.33880.83460.01200.69220.5827
      Proposed method7.747265.43860.07362.12780.82070.01460.68010.5896
    • Table 2. RoadScene dataset fusion results

      View table

      Table 2. RoadScene dataset fusion results

      MethodENSDSFMIFMIVIFQAB/FSSIM
      DIF6.896637.25330.09351.13990.46540.01870.73290.7673
      ReCoNet7.369453.11480.12521.10380.62900.01970.70730.6855
      RFNet6.980039.61730.14390.84990.45410.02560.71980.7729
      IFNet6.653631.72420.09101.14400.39250.01860.82800.8038
      FusionNet6.034322.06950.07510.75861.11630.01390.29890.4418
      DeepFuse6.931836.90140.10081.04430.48500.01940.83260.7575
      TFNet7.031041.96490.15871.13550.55750.02500.78970.7212
      Proposed method7.447150.75090.16231.93990.43500.02650.82830.7825
    • Table 3. TNO dataset fusion results

      View table

      Table 3. TNO dataset fusion results

      MethodENSDSFMIFMIVIFQAB/FSSIM
      DIF6.980355.59370.12880.86020.54250.02480.22280.7288
      ReCoNet7.179845.65170.18661.53450.73130.03120.20190.6344
      RFNet7.042557.52050.19180.79930.64790.03100.24410.6760
      IFNet6.600933.81090.10770.87770.53250.02260.33460.7337
      FusionNet4.460346.57390.02160.81061.57450.00560.23820.2128
      DeepFuse6.678427.66600.12330.66940.49710.02480.39680.7515
      TFNet6.570744.21570.15211.02340.69390.02780.33050.6531
      Proposed method7.211363.17790.18810.87800.85520.03180.33630.7624
    • Table 4. Ablation experiments of DWCL

      View table

      Table 4. Ablation experiments of DWCL

      MethodENSDSFMIFMIVIFQAB/FSSIM
      w/o DWCL4.325138.62480.16250.83610.81650.02650.26850.5484
      Cosine loss6.525246.68180.18251.15480.89930.02410.28870.6582
      Proposed method7.211363.17790.18810.87800.85520.03180.33630.7624
    • Table 5. Ablation experiment of double branch

      View table

      Table 5. Ablation experiment of double branch

      ModelENSDSFMIFMIVIFQAB/FSSIM
      Base encoder6.616443.68420.14650.65150.61500.03100.21650.7156
      Detail encoder6.356147.68450.16530.71540.72650.02850.28650.6894
      Proposed model7.211363.17790.18810.87800.85520.03180.33630.7624
    Tools

    Get Citation

    Copy Citation Text

    Linglin Bao, Pengge Ma, Long Wang, Jinwang Qian, Zhaoyu Liu, Qiuchun Jin. Dynamic Contrast Dual-Branch Feature Decomposition Network for Infrared-Visible Image Fusion[J]. Laser & Optoelectronics Progress, 2025, 62(16): 1637005

    Download Citation

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

    Category: Digital Image Processing

    Received: Jan. 15, 2025

    Accepted: Mar. 17, 2025

    Published Online: Aug. 11, 2025

    The Author Email: Pengge Ma (mapenge@163.com)

    DOI:10.3788/LOP250523

    CSTR:32186.14.LOP250523

    Topics