• Acta Optica Sinica
  • Vol. 40, Issue 2, 0210003 (2020)
Yong Chen*, Hongguang Guo, and Yapeng Ai
Author Affiliations
  • School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou, Gansu 730070, China
  • show less
    DOI: 10.3788/AOS202040.0210003 Cite this Article Set citation alerts
    Yong Chen, Hongguang Guo, Yapeng Ai. Single Image Dehazing of Multiscale Deep-Learning Based on Dual-Domain Decomposition[J]. Acta Optica Sinica, 2020, 40(2): 0210003 Copy Citation Text show less
    Physical model of atmospheric scattering
    Fig. 1. Physical model of atmospheric scattering
    Schematic of single image dehazing model of multiscale deep-learning based on dual-domain decomposition
    Fig. 2. Schematic of single image dehazing model of multiscale deep-learning based on dual-domain decomposition
    Original image and its high- and low-frequency sub-images. (a) Original image; (b) high-frequency sub-image; (c) low-frequency sub-image
    Fig. 3. Original image and its high- and low-frequency sub-images. (a) Original image; (b) high-frequency sub-image; (c) low-frequency sub-image
    Comparison of activation functions. (a) ReLU function; (b) PReLU function
    Fig. 4. Comparison of activation functions. (a) ReLU function; (b) PReLU function
    Schematic of single image dehazing algorithm of multiscale deep-learning based on dual-domain decomposition
    Fig. 5. Schematic of single image dehazing algorithm of multiscale deep-learning based on dual-domain decomposition
    Outdoor hazy images and corresponding scene transmission labels after depth map preprocessing
    Fig. 6. Outdoor hazy images and corresponding scene transmission labels after depth map preprocessing
    Experimental results of synthetic hazy images processed by different methods. (a) Synthetic hazy images; (b) stardard dehazed images; (c) method in Ref. [6]; (d) method in Ref. [9]; (e) method in Ref. [11]; (f) method in Ref. [12]; (g) method in Ref. [13]; (h) proposed method
    Fig. 7. Experimental results of synthetic hazy images processed by different methods. (a) Synthetic hazy images; (b) stardard dehazed images; (c) method in Ref. [6]; (d) method in Ref. [9]; (e) method in Ref. [11]; (f) method in Ref. [12]; (g) method in Ref. [13]; (h) proposed method
    Experimental results of real natural hazy images processed by different methods. (a) Hazy images; (b) method in Ref. [6]; (c) method in Ref. [9]; (d) method in Ref. [11]; (e) method in Ref. [12]; (f) method in Ref. [13]; (g) proposed method
    Fig. 8. Experimental results of real natural hazy images processed by different methods. (a) Hazy images; (b) method in Ref. [6]; (c) method in Ref. [9]; (d) method in Ref. [11]; (e) method in Ref. [12]; (f) method in Ref. [13]; (g) proposed method
    Image numberMethod in Ref. [6]Method in Ref. [9]Method in Ref. [11]Method in Ref. [12]Method in Ref. [13]Proposed method
    PSNR /dBSSIM /%PSNR /dBSSIM /%PSNR /dBSSIM /%PSNR /dBSSIM /%PSNR /dBSSIM /%PSNR /dBSSIM /%
    114.615075.1219.703777.7316.034977.6715.924779.2314.303870.2917.086582.55
    219.452178.4720.580480.9021.185481.0520.803082.6318.555576.6021.799188.42
    323.837486.2617.149346.2021.326385.7922.417280.2421.326385.7924.795391.15
    416.047776.4017.851981.2119.211085.3817.624579.1717.521777.9122.755087.69
    520.910888.3718.702077.0422.208091.7421.681290.6820.966489.0222.795492.41
    Table 1. Analysis of experimental results of synthetic hazy images processed by different methods
    Image numberMethod in Ref.[6]Method in Ref.[9]Method in Ref.[11]Method in Ref.[12]Method in Ref.[13]Proposed method
    IEAGIEAGIEAGIEAGIEAGIEAG
    17.447010.737.559915.487.789310.047.542011.587.621711.017.799117.96
    27.24015.187.10038.847.39325.647.35506.277.31816.857.530210.05
    37.43106.677.14989.457.35986.047.17666.307.15716.677.724310.46
    47.25024.336.84306.617.38764.297.56894.887.58255.857.61409.99
    57.567717.227.695520.557.602117.217.609516.487.705618.087.774523.61
    67.75566.797.493010.037.34475.247.29576.487.20505.997.98627.75
    77.04283.986.24228.797.43324.077.15304.177.17414.737.60256.36
    87.27607.237.719112.377.75298.197.71598.727.65228.977.877912.99
    Table 2. Analysis of experimental results of real hazy images processed by different methods
    Yong Chen, Hongguang Guo, Yapeng Ai. Single Image Dehazing of Multiscale Deep-Learning Based on Dual-Domain Decomposition[J]. Acta Optica Sinica, 2020, 40(2): 0210003
    Download Citation