Infrared and Laser Engineering, Volume. 49, Issue 5, 20200015(2020)

Multi-input fusion adversarial network for underwater image enhancement

Lin Sen1...2,3,*, Liu Shiben1, and Tang Yandong23 |Show fewer author(s)
Author Affiliations
  • 1[in Chinese]
  • 2[in Chinese]
  • 3[in Chinese]
  • show less
    Figures & Tables(10)
    Preprocessing graph
    Convolution and deconvolution
    Network structure diagram
    ReLu comparison experiment
    Skip connection experiment. (a) Original drawing; (b) No residual connection; (c) Residual connection; (d) Residual connection and one layer of convolution
    Experimental of color restoration. (a) Origianl;(b) Standard color card;(c) LAB;(d) DUIENet;(e) DCP;(f) DehazeNet and HWD;(g) MFGAN
    Experimental result . (a) Original;(b) DCP;(c) LAB;(d) CLAHE;(e) DehazeNet and HWD;(f) DUIENet;(g) UGAN;(h) MFGAN
    Experimental results of feature matching
    • Table 1. Quantitative results in terms of UCIQE

      View table
      View in Article

      Table 1. Quantitative results in terms of UCIQE

      ImageDCPLABCLAHEDehazeNet and HWDDUIENetUGANMFGAN
      10.676 20.600 80.645 50.669 70.630 30.690 60.676 6
      20.636 30.576 90.615 80.613 80.577 20.619 60.641 0
      30.595 10.466 60.461 60.546 90.441 20.561 60.637 7
      40.598 50.569 90.605 10.653 30.566 80.601 50.624 1
      50.590 30.546 80.577 70.580 70.593 80.577 70.624 4
      60.638 80.568 50.592 60.639 10.568 10.592 60.640 3
      Average0.622 50.554 90.583 10.617 20.562 90.607 30.639 9
    • Table 2. Quantitative results in terms of NIQE

      View table
      View in Article

      Table 2. Quantitative results in terms of NIQE

      ImageDCPLABCLAHEDehazeNet and HWDDUIENetUGANMFGAN
      13.559 53.850 23.909 84.715 64.700 64.663 03.614 4
      23.581 43.547 53.411 64.873 44.901 13.525 33.442 2
      36.457 16.047 45.804 98.510 46.034 83.327 04.548 6
      43.585 53.712 13.875 23.993 44.150 34.562 43.359 9
      54.050 14.065 93.904 63.868 54.507 85.097 93.745 3
      63.226 13.400 23.497 23.580 73.061 85.546 03.455 4
      Average4.076 64.058 94.067 24.923 64.559 44.453 63.727 3
    Tools

    Get Citation

    Copy Citation Text

    Lin Sen, Liu Shiben, Tang Yandong. Multi-input fusion adversarial network for underwater image enhancement[J]. Infrared and Laser Engineering, 2020, 49(5): 20200015

    Download Citation

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

    Category:

    Received: Feb. 14, 2020

    Accepted: Mar. 5, 2020

    Published Online: Sep. 22, 2020

    The Author Email: Sen Lin (lin_sen6@126.com)

    DOI:10.3788/irla20200015

    Topics