• Acta Photonica Sinica
  • Vol. 51, Issue 2, 0210004 (2022)
Yan YANG*, Jinlong ZHANG, and Xiaozhen LIANG
Author Affiliations
  • School of Electronic and Information Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China
  • show less
    DOI: 10.3788/gzxb20225102.0210004 Cite this Article
    Yan YANG, Jinlong ZHANG, Xiaozhen LIANG. End-to-end Image Dehazing Based on Ladder Network and Cross Fusion[J]. Acta Photonica Sinica, 2022, 51(2): 0210004 Copy Citation Text show less
    The network structure of this article
    Fig. 1. The network structure of this article
    Feature extraction network
    Fig. 2. Feature extraction network
    Hazy image and its feature map
    Fig. 3. Hazy image and its feature map
    Attention cross fusion module
    Fig. 4. Attention cross fusion module
    Image reconstruction
    Fig. 5. Image reconstruction
    Real hazy image feature and restoration schematic diagram
    Fig. 6. Real hazy image feature and restoration schematic diagram
    Experimental results of real hazy images
    Fig. 7. Experimental results of real hazy images
    Experimental results of the Reside dataset
    Fig. 8. Experimental results of the Reside dataset
    Experimental results of the NYU dataset
    Fig. 9. Experimental results of the NYU dataset
    Experimental results of the Middlebury dataset
    Fig. 10. Experimental results of the Middlebury dataset
    Partial enlarged schematic
    Fig. 11. Partial enlarged schematic
    Comparison of ablation experiments
    Fig. 12. Comparison of ablation experiments
    He'sWang'sCai'sLi'sRen'sQin'sOurs
    e0.328 80.265 70.216 60.272 10.318 60.243 10.376 8
    r1.380 51.274 31.333 61.302 21.440 11.218 71.449 3
    δ0.065 40.008 40.223 90.014 30.044 20.016 70.012 0
    HCC0.193 00.143 50.134 70.078 60.106 40.151 50.194 2
    Table 1. Comparison of real hazy image indicators of various algorithms
    He'sWang'sCai'sLi'sRen'sQin'sOurs
    PSNR15.8816.7820.1617.4716.8820.7621.47
    SSIM0.966 80.927 80.964 90.945 40.937 50.968 40.967 7
    Table 2. RESIDE test set indicators of every algorithm
    He'sWang'sCai'sLi'sRen'sQin'sOurs
    PSNR16.7117.4220.4819.0618.2720.9621.24
    SSIM0.927 00.893 50.912 90.904 90.901 60.923 80.928 9
    Table 3. NYU test set indicators of every algorithm
    He'sWang'sCai'sLi'sRen'sQin'sOurs
    PSNR15.2416.6919.8617.8917.2520.0820.46
    SSIM0.948 90.903 80.927 00.912 60.910 90.951 50.950 6
    Table 4. Middlebury test set indicators of every algorithm
    DFCFN-ACFACF
    PSNR18.5118.5519.2521.47
    SSIM0.943 70.960 90.971 50.987 7
    Table 5. Objective indicators of ablation experiment
    Yan YANG, Jinlong ZHANG, Xiaozhen LIANG. End-to-end Image Dehazing Based on Ladder Network and Cross Fusion[J]. Acta Photonica Sinica, 2022, 51(2): 0210004
    Download Citation