• Acta Photonica Sinica
  • Vol. 49, Issue 6, 0610001 (2020)
Wen XIAO1, Jie LI1, Feng PAN1、*, and Shuang ZHAO2
Author Affiliations
  • 1School of Instrumentation and Optoelectronics Engineering, Beihang University, Beijing 100191, China
  • 261206 Troops, Chinese People's Liberation Army, Beijing 100042, China
  • show less
    DOI: 10.3788/gzxb20204906.0610001 Cite this Article
    Wen XIAO, Jie LI, Feng PAN, Shuang ZHAO. Super-resolution in Digital Holographic Phase Cell Image Based on USENet[J]. Acta Photonica Sinica, 2020, 49(6): 0610001 Copy Citation Text show less
    Overall USENet structure
    Fig. 1. Overall USENet structure
    The general view of convolution layers for features extraction
    Fig. 2. The general view of convolution layers for features extraction
    The feature maps of weights calibration layer
    Fig. 3. The feature maps of weights calibration layer
    The training and validation process
    Fig. 4. The training and validation process
    Reconstruction steps on validation set
    Fig. 5. Reconstruction steps on validation set
    Fitting curves plotted by assessment scores from the input and output of 300 samples in validation set
    Fig. 6. Fitting curves plotted by assessment scores from the input and output of 300 samples in validation set
    Fitting curves plotted by assessment scores from the input and output of 300 samples in validation set
    Fig. 7. Fitting curves plotted by assessment scores from the input and output of 300 samples in validation set
    Sample image in spectral domain
    Fig. 8. Sample image in spectral domain
    Isopleth and three-dimensional morphology of cells
    Fig. 9. Isopleth and three-dimensional morphology of cells
    Fitting curves of SSIM performance in the ROI by the rebuilding of validation set
    Fig. 10. Fitting curves of SSIM performance in the ROI by the rebuilding of validation set
    Detailed internal structure of USENet
    Fig. 11. Detailed internal structure of USENet
    IDαProportion of L1βProportion of MSEγProportionof L2
    111.0000.0000.00
    20.10.9010.1000.00
    30.030.7010.3000.00
    40.0250.6510.3500.00
    50.020.6010.4000.00
    60.0150.5510.4500.00
    70.0120.5010.5000.00
    80.010.4510.5500.00
    90.0080.4010.6000.00
    100.0050.3510.6500.00
    110.0030.2010.8000.00
    1200.0011.0000.00
    1300.0000.0011.00
    140.0250.650.90.300.0020.05
    150.0200.5510.400.0020.05
    Table 1. Approaches with 15 different combinations of weighted loss factors
    IDSSIM_OPTSSIM_AVGMSE_OPT×10-3MSE_AVG×10-3PSNR_OPTPSNR_AVG
    10.940 10.937 70.9891.01842.62842.516
    20.936 50.930 11.0271.13142.30741.992
    30.939 50.935 00.9081.03543.06142.461
    40.942 70.93980.8470.86243.40343.308
    50.941 10.93690.8600.89143.32943.151
    60.941 60.937 10.8450.85843.41043.329
    70.939 50.937 70.8410.85843.43643.332
    80.940 30.935 40.8640.88243.29343.175
    90.939 70.936 60.8370.84643.45743.406
    100.937 00.935 10.9280.94742.95142.172
    110.933 70.927 81.0531.22441.91841.327
    120.928 20.919 20.9120.99042.94842.689
    130.928 50.920 10.9941.04442.63442.007
    140.940 90.931 80.8941.15243.11742.131
    150.937 60.928 60.9131.15443.02742.171
    Table 2. The performances of net in convergence interval on validation set with 15 different loss functions
    ModelGlobal (SSIM)ROI(SSIM)
    USENet0.942 70.970 3
    Reference0.940 90.965 5
    Table 3. The influence of ROI rebuilding by the calibration layer
    Wen XIAO, Jie LI, Feng PAN, Shuang ZHAO. Super-resolution in Digital Holographic Phase Cell Image Based on USENet[J]. Acta Photonica Sinica, 2020, 49(6): 0610001
    Download Citation