Optics and Precision Engineering, Volume. 32, Issue 12, 1915(2024)

Wavelet dehazeformer network for road traffic image dehazing method

Ping XIA1...2, Ziyi LI1,2, Bangjun LEI1,2,*, Yudie WANG1,2, and Tinglong TANG12 |Show fewer author(s)
Author Affiliations
  • 1Hubei Key Laboratory of Intelligent Vision based Monitoring for Hydroelectric Engineering, Three Gorges University, Yichang443002, China
  • 2College of Computer and Information Technology, Three Gorges University, Yichang44300, China
  • show less
    Figures & Tables(15)
    Model of this paper
    SKFF Model
    Intermediate feature layer model
    CGA model
    CGAFusion
    Jump connection processing effects
    Synthetic Fog Image Generation
    Shows the dehazing results of different methods. In (a1)-(a2), foggy images from the Foggy_Cityscapes dataset are presented. In (a3)-(a4), foggy images from the 4K-HAZE dataset are displayed. In (a5)-(a6), foggy images from the DKITTI dataset are depicted. Subfigures (b) through (i) represent dehazing results using various methods: (b) Dark Channel Prior, (c) AOD-Net, (d) Wavelet-Net, (e) FFA-Net, (f) EPDN, (g) DehazeFormer, (h) the proposed method, and (i) the ground truth image
    Enlarged View of the Red Region in the Fifth Dehazed Image of Figure 7; Among them, (ak) foggy image; (bk) the result of dark channel dehazing; (ck) the result of AOD-Net dehazing; (dk) the result of Wavelet-Net dehazing; (ek) the result of FFA-Net dehazing; (fk) the result of EPDN dehazing; (gk) the result of DehazeFormer dehazing; (hk) the result of the proposed method dehazing; (ik) the real image ; k=2,4,5
    Comparison of PSNR and SSIM for Different Dehazing Algorithms
    Subjective Comparison of Ablation Experiments
    • Table 1. Comparative analysis of performance on the foggy_cityscapes

      View table
      View in Article

      Table 1. Comparative analysis of performance on the foggy_cityscapes

      PSNRSSIMMSEEntropy
      DCP915.890.678 21 874.866.70
      AOD-Net3017.120.787 21 426.616.23
      Wavelet-Net1517.300.745 31 416.236.31
      FFA-Net2023.200.911 7393.756.66
      EPDN3123.610.901 6329.176.80
      DehazeFormer1726.240.940 4198.306.81
      本文方法27.560.955 8168.856.87
    • Table 2. Comparative Analysis of Performance on the 4K-HAZE Dataset

      View table
      View in Article

      Table 2. Comparative Analysis of Performance on the 4K-HAZE Dataset

      PSNRSSIMMSEEntropy
      DCP915.370.677 33 243.337.20
      AOD-Net3013.030.117 54 100.882.89
      Wavelet-Net1517.850.662 61 111.796.14
      FFA-Net2019.360.803 3993.676.77
      EPDN3123.160.820 1395.716.73
      DehazeFormer1721.830.837 7678.696.61
      本文方法24.220.902 5367.177.09
    • Table 3. Comparative analysis of performance on the DKITTI Dataset

      View table
      View in Article

      Table 3. Comparative analysis of performance on the DKITTI Dataset

      PSNRSSIMMSEEntropy
      DCP913.060.650 24 037.397.26
      AOD-Net3012.780.591 13 922.137.28
      Wavelet-Net1513.630.466 22 974.547.29
      FFA-Net2021.920.780 6483.997.31
      EPDN3120.010.725 5708.957.27
      DehazeFormer1723.960.811 4306.637.12
      本文方法24.520.826 0283.487.12
    • Table 4. Ablation study

      View table
      View in Article

      Table 4. Ablation study

      BaselineDSCWResidualSKFFPSNRSSIMMSEEntropy
      Dehazeformer26.200.932 6202.916.79
      26.270.942 9207.036.83
      26.260.945 8200.546.82
      26.490.944 6188.106.84
      26.940.942 9186.16.82
      27.560.955 8168.856.87
    Tools

    Get Citation

    Copy Citation Text

    Ping XIA, Ziyi LI, Bangjun LEI, Yudie WANG, Tinglong TANG. Wavelet dehazeformer network for road traffic image dehazing method[J]. Optics and Precision Engineering, 2024, 32(12): 1915

    Download Citation

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

    Category:

    Received: Mar. 5, 2024

    Accepted: --

    Published Online: Aug. 28, 2024

    The Author Email: LEI Bangjun (Bangjun.Lei@ieee.org)

    DOI:10.37188/OPE.20243212.1915

    Topics