• Laser & Optoelectronics Progress
  • Vol. 57, Issue 22, 221018 (2020)
Jianwang Gan1, Yun Sha1、*, and Guoying Zhang2
Author Affiliations
  • 1School of Information Engineering, Beijing Institute of Petrochemical Technology, Beijing 102617, China
  • 2School of Mechanical Electronic & Information Engineering, China University of Mining and Technology (Beijing), Beijing 100083, China;
  • show less
    DOI: 10.3788/LOP57.221018 Cite this Article Set citation alerts
    Jianwang Gan, Yun Sha, Guoying Zhang. Image Denoising Based on Asymmetric Convolutional Neural Networks[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221018 Copy Citation Text show less
    Partially synthesized noise images.
    Fig. 1. Partially synthesized noise images.
    CBDNet model
    Fig. 2. CBDNet model
    CBDNet denoising subnet architecture
    Fig. 3. CBDNet denoising subnet architecture
    UNet++ architecture
    Fig. 4. UNet++ architecture
    Improved network model
    Fig. 5. Improved network model
    Denoising results of image 1. (a) Original image; (b) noise image; (c) CBDNet; (d) improved CBDNet
    Fig. 6. Denoising results of image 1. (a) Original image; (b) noise image; (c) CBDNet; (d) improved CBDNet
    Denoising results of image 2. (a) Original image; (b) noise image; (c) CBDNet; (d) improved CBDNet
    Fig. 7. Denoising results of image 2. (a) Original image; (b) noise image; (c) CBDNet; (d) improved CBDNet
    Denoising results of image 3. (a) Original image; (b) noise image; (c) CBDNet; (d) improved CBDNet
    Fig. 8. Denoising results of image 3. (a) Original image; (b) noise image; (c) CBDNet; (d) improved CBDNet
    No.CBDNetImproved CBDNet
    135.76972035.880955
    231.46179430.128640
    336.20212036.103760
    433.20525733.806810
    535.03499234.901047
    638.66932740.513270
    737.77080537.781620
    834.80051834.688103
    934.61019534.964275
    1034.23290334.225975
    Mean value35.17576235.299446
    Table 1. PSNR of 10 images
    No.CBDNetImproved CBDNet
    10.941740700.94231653
    20.963752570.96220344
    30.972951230.97360283
    40.975487050.97699260
    50.977546300.97723440
    60.979628560.97987250
    70.982511040.98210240
    80.985826700.98583820
    90.969948600.97106440
    100.979355630.97972226
    Mean value0.972874800.97309494
    Table 2. SSIM of 10 images
    TypeBM3DCBDNetImproved CBDNet
    PSNR35.0035.20035.400
    SSIM0.860.9720.973
    Table 3. Mean of PSNR and SSIM
    Jianwang Gan, Yun Sha, Guoying Zhang. Image Denoising Based on Asymmetric Convolutional Neural Networks[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221018
    Download Citation