Photonics Research, Volume. 12, Issue 8, 1640(2024)

Robust polarimetric dehazing algorithm based on low-rank approximation and multiple virtual-exposure fusion

Yifu Zhou1,2, Hanyue Wei1,2, Jian Liang1,2, Feiya Ma1,2, Rui Yang1,2, Liyong Ren1,2,3、*, and Xuelong Li4,5
Author Affiliations
  • 1School of Physics and Information Technology, Shaanxi Normal University, Xi’an 710119, China
  • 2Xi’an Key Laboratory of Optical Information Manipulation and Augmentation (OMA), Xi’an 710119, China
  • 3Robust (Xixian New Area) Opto-Electro Technologies Co., Ltd., Xi’an 712000, China
  • 4Institute of Artificial Intelligence (TeleAI), China Telecom Corp Ltd., Beijing 100033, China
  • 5e-mail: xuelong_li@ieee.org
  • show less
    Figures & Tables(19)
    Diagram of the optical degradation model.
    (a) Flow chart of the construction of M; (b) flow chart of the calculation of Iu.
    Flow chart of the exposure fusion.
    Calculating four saturation threshold values of an HDR dehazed image.
    Generating a bracketed exposure sequence and fusing them into an SDR dehazed image.
    Flow chart of the proposed polarimetric dehazing algorithm.
    Hazy images and corresponding dehazed results.
    (a) Sky area of images of the experimental group (c); (b) pylon in images of the experimental group (a); (c) pylon in images of the experimental group (e).
    (a) Hazy image; (b) estimated DoLP by the low-rank approximation method; (c) estimated DoLP by the Stokes method; (d) dehazed image based on DoLP shown in (b); (e) dehazed image based on DoLP shown in (c); (f)–(i) local views corresponding to (b)–(e), respectively.
    (a) Estimated intensity map of scattering light based on estimated DoLP by low-rank approximation method; (b) estimated intensity map of scattering light based on estimated DoLP by the Stokes method; (c), (d) local views corresponding to (a), (b), respectively; (e) intensity curves of pixels on the line of y=400 in (a) and (b).
    (a) Hazy image; (b) HDR dehazed image; (c) SDR dehazed image by the MVEF; (d)–(g) SDR dehazed images by the gamma correction with different values of γ.
    Standard variances of images shown in Fig. 11.
    Hazy images and corresponding dehazed images by the proposed method and our previous method [17].
    (a) Hazy image; (b) denoised intensity map of scattering light by the low-rank approximation method; (c) denoised intensity map of scattering light by the low-pass filtering method; (d) dehazed image based on DoLP shown in (b); (e) dehazed image based on DoLP shown in (c); (f), (g) local views corresponding to (d) and (e), respectively.
    Local areas in hazy images and corresponding dehazed images shown in Fig. 13.
    • Table 1. Standard Variances of Dehazing Results

      View table
      View in Article

      Table 1. Standard Variances of Dehazing Results

      GroupOriginal ImageProposedProposed without MVEFDark Channel Prior MethodSchechner’s Method
      Fig. 7(a)43.9970.8470.6653.5851.46
      Fig. 7(b)17.6156.6256.4540.8044.47
      Fig. 7(c)31.4865.8064.7245.2380.87
      Fig. 7(d)20.8241.3241.6935.2826.97
      Fig. 7(e)28.0771.9955.7054.1062.39
      Fig. 7(f)16.0956.2534.9026.0635.63
      Fig. 7(g)21.5554.4540.1952.5035.28
    • Table 2. Information Entropies of Dehazing Results

      View table
      View in Article

      Table 2. Information Entropies of Dehazing Results

      GroupOriginal ImageProposedProposed without MVEFDark Channel Prior MethodSchechner’s Method
      Fig. 7(a)7.247.817.947.677.50
      Fig. 7(b)6.147.607.567.237.40
      Fig. 7(c)6.897.607.777.417.41
      Fig. 7(d)6.307.347.236.986.74
      Fig. 7(e)6.677.586.927.637.58
      Fig. 7(f)5.797.716.616.366.89
      Fig. 7(g)6.417.626.697.577.10
    • Table 3. NIQE Scores of Dehazing Results

      View table
      View in Article

      Table 3. NIQE Scores of Dehazing Results

      GroupOriginal ImageProposedProposed without MVEFDark Channel Prior MethodSchechner’s Method
      Fig. 7(a)5.303.443.473.813.60
      Fig. 7(b)6.352.853.003.674.06
      Fig. 7(c)5.993.613.523.653.71
      Fig. 7(d)7.984.364.744.634.99
      Fig. 7(e)7.293.733.554.284.58
      Fig. 7(f)6.864.044.594.054.46
      Fig. 7(g)6.434.104.673.744.33
    • Table 4. PIQE Scores of Dehazing Results

      View table
      View in Article

      Table 4. PIQE Scores of Dehazing Results

      GroupOriginal ImageProposedProposed without MVEFDark Channel Prior MethodSchechner’s Method
      Fig. 7(a)36.6810.7119.1424.427.95
      Fig. 7(b)43.0117.9319.1626.7519.24
      Fig. 7(c)33.4213.6814.0626.2630.60
      Fig. 7(d)28.4016.597.4626.0523.03
      Fig. 7(e)41.9624.2519.0637.4729.20
      Fig. 7(f)34.8017.859.729.224.46
      Fig. 7(g)25.5310.3911.8213.0012.94
    Tools

    Get Citation

    Copy Citation Text

    Yifu Zhou, Hanyue Wei, Jian Liang, Feiya Ma, Rui Yang, Liyong Ren, Xuelong Li, "Robust polarimetric dehazing algorithm based on low-rank approximation and multiple virtual-exposure fusion," Photonics Res. 12, 1640 (2024)

    Download Citation

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

    Category: Image Processing and Image Analysis

    Received: Feb. 26, 2024

    Accepted: May. 20, 2024

    Published Online: Jul. 25, 2024

    The Author Email: Liyong Ren (renliy@snnu.edu.cn)

    DOI:10.1364/PRJ.522370

    Topics