• Infrared and Laser Engineering
  • Vol. 54, Issue 5, 20240570 (2025)
Shuai LIU1, Mingjun WANG1,2,3, and Yiming ZHOU1
Author Affiliations
  • 1School of Automation and Information Engineering, Xi’an University of Technology, Xi’an 710048, China
  • 2Xi'an Key Laboratory of Wireless Optical Communication and Network Research, Xi’an 710048, China
  • 3School of Physics and Telecommunications Engineering, Shaanxi University of Technology, Hanzhong 723001, China
  • show less
    DOI: 10.3788/IRLA20240570 Cite this Article
    Shuai LIU, Mingjun WANG, Yiming ZHOU. Non-Line-of-Sight imaging based on improved CNN-Transformer[J]. Infrared and Laser Engineering, 2025, 54(5): 20240570 Copy Citation Text show less
    NLOS imaging system schematic diagram
    Fig. 1. NLOS imaging system schematic diagram
    Improved CNN-Transformer NLOS imaging flowchart
    Fig. 2. Improved CNN-Transformer NLOS imaging flowchart
    CNN-Transformer block
    Fig. 3. CNN-Transformer block
    Reconstruction results of seen and unseen test sets on simulated dataset
    Fig. 4. Reconstruction results of seen and unseen test sets on simulated dataset
    Qualitative comparison of MobileNet and ResNet on real-world data
    Fig. 5. Qualitative comparison of MobileNet and ResNet on real-world data
    Qualitative comparison of different parameter combinations on real-world data
    Fig. 6. Qualitative comparison of different parameter combinations on real-world data
    Imaging results of reconstructed hidden scenes from the public real-world dataset [7]
    Fig. 7. Imaging results of reconstructed hidden scenes from the public real-world dataset [7]
    DataMethodsIntensityDepth
    PSNR/dBSSIM/dBRMSEMAD
    SeenFBP19.960.18460.70530.6694
    LCT19.780.44770.66940.6321
    FK22.170.42570.71560.6846
    RSD23.110.79960.55580.5332
    LFE26.900.86610.07690.0455
    Ours28.560.93380.06060.0271
    UnseenFBP17.810.21140.69860.6479
    LCT18.540.49620.66040.6152
    FK19.580.41510.73350.6938
    RSD19.920.77290.58960.5526
    LFE23.400.81000.12200.0561
    Ours24.540.87230.11470.0472
    Table 1. Quantitative comparison of different methods on seen and unseen test sets
    MethodIntensityDepth
    PSNR/dBSSIM/dBRMSEMAD
    ResNet29.546660.953690.034620.03368
    MobileNet29.829770.956100.033680.02736
    Table 2. Quantitative comparison between MobileNet and ResNet on seen test sets
    GroupIntensityDepth
    PSNR/dBSSIM/dBRMSEMAD
    129.518990.951060.034760.00724
    229.531310.953480.034580.00771
    329.546660.956100.034620.00699
    429.577230.959970.034020.00604
    529.803040.963120.033790.00515
    Table 3. Quantitative comparison of different parame-ter combinations on seen test sets
    Shuai LIU, Mingjun WANG, Yiming ZHOU. Non-Line-of-Sight imaging based on improved CNN-Transformer[J]. Infrared and Laser Engineering, 2025, 54(5): 20240570
    Download Citation