Acta Optica Sinica, Volume. 39, Issue 11, 1110002(2019)

Image Dehazing Algorithm Based on Convolutional Neural Network and Dynamic Ambient Light

Jieping Liu*, Yezhang Yang, Minyuan Chen, and Lihong Ma
Author Affiliations
  • School of Electronic and Information Engineering, South China University of Technology, Guangzhou, Guangdong 510641, China
  • show less
    Figures & Tables(12)
    Framework of the proposed algorithm
    Coarse segmentation of ambient light images. (a) Hazy images; (b) histograms of V channel; (c) regional segmentation
    Dynamic ambient light. (a) Hazy images; (b) rough ambient light maps; (c) refined ambient light maps
    Examples of training set. (a) Real hazy images; (b) transmittance images; (c) paired training samples
    Structure of TEN
    Estimation and refinement of transmittance. (a) Hazy images; (b) transmittance estimated by TEN; (c) refined transimittance
    Comparison of transmittance estimation effects. (a) Hazy images; (b) transmittance estimated by TEN2; (c) transmittance estimated by TEN1; (d) restored results of TEN2; (e) restored results of TEN1
    Restored effects of global and dynamic ambient light. (a) Hazy images; (b) results restored by global atmospheric light; (c) results restored by dynamic ambient light
    Dehazing results of different algorithms. (a) Hazy images; (b) method in Ref. [4-5]; (c) method in Ref. [6]; (d) method in Ref. [9]; (e) method in Ref. [10]; (f) proposed algorithm
    • Table 1. Comparison of average objective indexes of dehazing images using the trained network with different training sets

      View table

      Table 1. Comparison of average objective indexes of dehazing images using the trained network with different training sets

      Training setStandard deviationAverage gradientInformation entropyEdge intensity
      Ref. [6]60.218.227.2159.26
      Ref. [9]59.238.057.2258.07
      Ref. [7]62.598.177.2258.80
      Ref. [18]64.029.557.2169.03
    • Table 2. Comparison of average objective indexes of dehazing images with different algorithms

      View table

      Table 2. Comparison of average objective indexes of dehazing images with different algorithms

      AlgorithmStandard deviationAverage gradientInformation entropyEdge intensity
      Method in Ref. [4-5]45.127.827.0257.07
      Method in Ref. [6]51.357.657.1455.50
      Method in Ref. [9]51.417.137.1651.34
      Method in Ref. [10]51.328.047.2057.82
      Proposed method64.029.557.2169.03
    • Table 3. Comparison of objective indicators of dehazing images with different algorithms

      View table

      Table 3. Comparison of objective indicators of dehazing images with different algorithms

      ImageAlgorithmStandard deviationAverage gradientInformation entropyEdge intensity
      Img1Method in Ref. [4-5]52.957.477.4058.53
      Method in Ref. [6]61.707.497.3459.37
      Method in Ref. [9]66.516.957.1955.70
      Method in Ref. [10]62.457.277.3458.09
      Proposed method76.289.007.5471.62
      Img2Method in Ref. [4-5]45.3313.027.4789.47
      Method in Ref. [6]49.4011.297.3376.85
      Method in Ref. [9]37.438.907.0962.69
      Method in Ref. [10]57.0213.057.5388.99
      Proposed method70.1115.257.63104.10
      Img3Method in Ref. [4-5]55.7412.987.4186.35
      Method in Ref. [6]55.6012.967.6186.28
      Method in Ref. [9]52.0810.857.5274.17
      Method in Ref. [10]67.5414.617.6297.78
      Proposed method77.6318.017.23119.85
      Img4Method in Ref. [4-5]49.3017.697.54119.87
      Method in Ref. [6]49.9615.827.56107.62
      Method in Ref. [9]56.7815.937.60106.44
      Method in Ref. [10]59.8918.477.75120.96
      Proposed method76.4722.087.60149.85
      Img5Method in Ref. [4-5]49.679.817.2765.23
      Method in Ref. [6]49.9810.227.2967.33
      Method in Ref. [9]50.899.407.2965.60
      Method in Ref. [10]54.459.897.2469.11
      Proposed method61.9512.117.3678.80
      Img6Method in Ref. [4-5]33.075.116.8235.96
      Method in Ref. [6]51.435.457.3538.38
      Method in Ref. [9]59.434.707.5734.38
      Method in Ref. [10]50.145.037.4036.77
      Proposed method72.296.757.6346.94
    Tools

    Get Citation

    Copy Citation Text

    Jieping Liu, Yezhang Yang, Minyuan Chen, Lihong Ma. Image Dehazing Algorithm Based on Convolutional Neural Network and Dynamic Ambient Light[J]. Acta Optica Sinica, 2019, 39(11): 1110002

    Download Citation

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

    Category: Image Processing

    Received: May. 31, 2019

    Accepted: Jul. 24, 2019

    Published Online: Nov. 6, 2019

    The Author Email: Liu Jieping (eeliujp@scut.edu.cn)

    DOI:10.3788/AOS201939.1110002

    Topics