Acta Optica Sinica, Volume. 43, Issue 5, 0518002(2023)

Large-Field Microscopic Imaging Method Based on Cycle Generative Adversarial Networks

Rao Fu1,2, Yu Fang1,2, Yong Yang4, Dong Xiang1,2, and Xiaojing Wu3、*
Author Affiliations
  • 1Institute of Modern Optics, Nankai University, Tianjin 300350, China
  • 2Tianjin Key Laboratory of Micro-Scale Optical Information Science and Technology, Tianjin 300350, China
  • 3Tianjin Union Medical Center, Tianjin 300121, China
  • 4Institute of Intelligent Sensing, Zhejiang Lab, Hangzhou 310013, Zhejiang, China
  • show less
    Figures & Tables(12)
    Model structure of GANs
    Model structure of Cycle-GANs
    Model illustration of Cycle-GANs
    Graded scaling of image resolution and prediction effect
    Fitting curve of relationship between objective evaluation index and image resolution reduction factor. (a) SSIM; (b) PSNR; (c) NMRSE
    Results of 25× high-resolution virtual images generated from low-resolution images of 10× resolution testing board
    Results of 25× HR generated from 10× LR
    Results of 25× HR generated from 4× LR
    • Table 1. Training parameters of image network model

      View table

      Table 1. Training parameters of image network model

      TypeDescription
      Image size(900,900,3)
      OptimizerAdam
      Learning rate0.0002
      Batch size1
      Epoch1000
      Decay epoch1000
      Crop size240
    • Table 2. Objective evaluation indexes of theory verification

      View table

      Table 2. Objective evaluation indexes of theory verification

      Image groupSSIMPSNR /dBNRMSE
      10.88926.8350.101
      20.83526.5120.109
      30.65524.6260.142
      40.63823.9710.152
      50.40719.5500.241
      60.39818.6310.265
    • Table 3. Objective evaluation indexes of 25× HR generated from 10× LR

      View table

      Table 3. Objective evaluation indexes of 25× HR generated from 10× LR

      Image groupSSIMPSNR /dBNRMSE
      10.68423.5230.131
      20.65723.0630.137
      30.60322.2790.138
    • Table 4. Objective evaluation indexes of 25× HR generated from 4× LR

      View table

      Table 4. Objective evaluation indexes of 25× HR generated from 4× LR

      Image groupSSIMPSNR /dBNRMSE
      10.59722.3220.180
      20.59622.2790.182
      30.58521.7970.184
    Tools

    Get Citation

    Copy Citation Text

    Rao Fu, Yu Fang, Yong Yang, Dong Xiang, Xiaojing Wu. Large-Field Microscopic Imaging Method Based on Cycle Generative Adversarial Networks[J]. Acta Optica Sinica, 2023, 43(5): 0518002

    Download Citation

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

    Category: Microscopy

    Received: Aug. 29, 2022

    Accepted: Oct. 14, 2022

    Published Online: Feb. 27, 2023

    The Author Email: Wu Xiaojing (xiaojingwu@nankai.edu.cn)

    DOI:10.3788/AOS221657

    Topics