Optics and Precision Engineering, Volume. 33, Issue 10, 1609(2025)

Multi-scale enhancement and color depth codec correction of flotation foam low illumination images

Lei SUN1,2, Qian TANG1, Yipeng LIAO1、*, Yuhua LIAO1, Zexi DONG1, and Jianjun HE3
Author Affiliations
  • 1College of Physics and Information Engineering, Fuzhou University, Fuzhou35008, China
  • 2School of Zhicheng College, Fuzhou University, Fuzhou35000, China
  • 3Fujian Jindong Mining Co. Ltd., Sanming65101,China
  • show less
    Figures & Tables(19)
    Overall framework diagram of the algorithm
    CDCC-Net and submodules architecture
    GSM and submodules architecture
    Multiscale enhancement and denoising process of V
    Enhancement effect of each method on three different working condition foam dataset
    Enhancement effect of each method on LZFF dataset
    Enhancement effect of each method on LRFF dataset
    Enhancement effect of each method on Low-Light dataset
    Denoising and edge detection effectiveness of each method on LZFF dataset
    Denoising and edge detection effectiveness of each method on Low-Light dataset
    Visual effect of each color space on LZFF dataset
    • Table 1. Comparison of the enhancement of each method on three different working condition foam datasets

      View table
      View in Article

      Table 1. Comparison of the enhancement of each method on three different working condition foam datasets

      MethodComentropyContrastResolutionTexture correlation/%PSNR/dB
      Low illumination foam image6.397 022.300 812.114 020.9324.043 1
      Method in Ref.[137.200 722.453 315.758 57.8214.582 2
      Method in Ref.[146.682 750.603 819.868 112.8715.011 4
      Method in Ref.[156.508 430.750 916.368 49.3024.706 1
      Method in Ref.[97.540 559.052 424.614 323.1325.867 0
      Proposed method7.683 161.431 625.677 824.7126.391 7
    • Table 2. Comparison of the performance of the methods on LZFF dataset

      View table
      View in Article

      Table 2. Comparison of the performance of the methods on LZFF dataset

      MethodPSNR/dB↑SSIM↑CIEDE↓
      Zero-DCE1616.8410.51024.440
      HDRNet179.9100.59421.338
      LCDPNet1810.9950.30421.211
      LA-Net1920.7270.5996.728
      RUAS2019.5770.6325.497
      MSRCR1010.8610.27821.387
      ZeroIG219.3180.15335.890
      GDNet2215.2460.78918.686
      LYT-Net2320.8130.75011.720
      Proposed method24.0900.8411.007
    • Table 3. Comparison of the performance of the methods on LRFF dataset

      View table
      View in Article

      Table 3. Comparison of the performance of the methods on LRFF dataset

      MethodPSNR/dB↑SSIM↑CIEDE↓
      Zero-DCE166.8760.29083.815
      HDRNet177.9560.39072.202
      LCDPNet185.8590.18191.877
      LA-Net1927.3480.8723.734
      RUAS209.2120.37262.047
      MSRCR109.9300.45151.305
      ZeroIG218.6600.36568.310
      GDNet2221.7290.8305.965
      LYT-Net2324.9220.8355.283
      Proposed method28.4610.9052.502
    • Table 4. Comparison of the performance of the methods on the Low-Light dataset

      View table
      View in Article

      Table 4. Comparison of the performance of the methods on the Low-Light dataset

      MethodPSNR/dB↑SSIM↑CIEDE↓
      Zero-DCE169.4060.35343.697
      HDRNet1714.7340.62245.445
      LCDPNet1810.2390.35753.697
      LA-Net1923.3870.86915.375
      RUAS2013.0670.68014.826
      MSRCR1014.2100.52318.681
      ZeroIG2118.8350.71213.147
      GDNet2223.7460.78417.131
      LYT-Net2216.5450.79914.804
      Proposed method21.6570.87210.088
    • Table 5. Denoising and edge detection performance statistics of each method on Low-Light dataset

      View table
      View in Article

      Table 5. Denoising and edge detection performance statistics of each method on Low-Light dataset

      Noise varianceIndexRUASLA-NetHDRNetMethod in Ref.[20Method in Ref.[14Method in Ref.[9Proposed method
      10%PSNR/dB14.191 122.608 812.455 222.812 722.903 923.271 323.874 8
      EPI/%0.551 90.662 70.569 30.332 30.189 50.647 20.690 9
      20%PSNR/dB14.186 022.520 112.291 722.780 221.903 322.294 322.684 1
      EPI/%0.547 80.664 60.568 10.337 40.194 10.635 80.675 7
      30%PSNR/dB14.130 022.333 511.275 121.782 620.909 421.392 621.524 2
      EPI/%0.543 00.660 80.556 10.520 20.588 10.661 80.663 5
    • Table 6. Denoising and edge detection performance statistics of each method on LZFF dataset

      View table
      View in Article

      Table 6. Denoising and edge detection performance statistics of each method on LZFF dataset

      Noise varianceIndexRUASLA-NetHDRNetMethod in Ref.[20Method in Ref.[14Method in Ref.[9Proposed method
      10%PSNR/dB14.191 122.608 812.455 222.812 722.903 923.271 323.874 8
      EPI/%0.551 90.662 70.569 30.332 30.189 50.647 20.690 9
      20%PSNR/dB14.186 022.520 112.291 722.780 221.903 322.294 322.684 1
      EPI/%0.547 80.664 60.568 10.337 40.194 10.635 80.675 7
      30%PSNR/dB14.130 022.333 511.275 121.782 620.909 421.392 621.524 2
      EPI/%0.543 00.660 80.556 10.520 20.588 10.661 80.663 5
    • Table 7. Ablation study of CDCC-Net modules on LZFF dataset

      View table
      View in Article

      Table 7. Ablation study of CDCC-Net modules on LZFF dataset

      BackboneGSMSimAPSNR/dB↑SSIM↑CIEDE↓
      22.4380.8153.572
      23.6960.8353.719
      22.6040.8253.644
      24.0900.8412.313
    • Table 8. Ablation study of color space on LZFF dataset

      View table
      View in Article

      Table 8. Ablation study of color space on LZFF dataset

      Color SpaceCDCC-NetProseed method of VPSNR/dB↑SSIM↑
      LAB22.2090.717
      21.5520.805
      22.3970.829
      LCH21.2740.824
      21.5190.784
      22.6810.837
      HIS23.8530.826
      23.0440.845
      24.9160.893
      HSV24.0900.841
      23.2810.857
      29.0620.930
    Tools

    Get Citation

    Copy Citation Text

    Lei SUN, Qian TANG, Yipeng LIAO, Yuhua LIAO, Zexi DONG, Jianjun HE. Multi-scale enhancement and color depth codec correction of flotation foam low illumination images[J]. Optics and Precision Engineering, 2025, 33(10): 1609

    Download Citation

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

    Category:

    Received: Dec. 17, 2024

    Accepted: --

    Published Online: Jul. 23, 2025

    The Author Email: Yipeng LIAO (fzu_lyp@163.com)

    DOI:10.37188/OPE.20253310.1609

    Topics