• Acta Photonica Sinica
  • Vol. 51, Issue 6, 0610002 (2022)
Xia WANG1、2、*, Xin ZHANG2, Gangcheng JIAO1, Ye YANG1, Hongchang CHENG1, and Bo YAN1
Author Affiliations
  • 1Science and Technology on Low-Light-Level Night Vision Laboratory,Xi'an 710065,China
  • 2Key Laboratory of Optoelectronic Imaging Technology and System,Ministry of Education,School of Optics and Photonics,Beijing Institute of Technology,Beijing 100081,China
  • show less
    DOI: 10.3788/gzxb20225106.0610002 Cite this Article
    Xia WANG, Xin ZHANG, Gangcheng JIAO, Ye YANG, Hongchang CHENG, Bo YAN. Dual Residual Attention Network for ICMOS Sensing Image[J]. Acta Photonica Sinica, 2022, 51(6): 0610002 Copy Citation Text show less
    The pipeline of ICMOS
    Fig. 1. The pipeline of ICMOS
    The overview of the method
    Fig. 2. The overview of the method
    The structure of the residual block
    Fig. 3. The structure of the residual block
    The structure of channel attention block
    Fig. 4. The structure of channel attention block
    Examples under different illumination in the datasets
    Fig. 5. Examples under different illumination in the datasets
    Denoising results of parrot image under 2×10-1 lx
    Fig. 6. Denoising results of parrot image under 2×10-1 lx
    Denoising results of butterfly image under 2×10-1 lx
    Fig. 7. Denoising results of butterfly image under 2×10-1 lx
    Denoising results of parrot image under 3×10-2 lx
    Fig. 8. Denoising results of parrot image under 3×10-2 lx
    Denoising results of butterfly image under 3×10-2 lx
    Fig. 9. Denoising results of butterfly image under 3×10-2 lx
    Denoising results of parrot image under 2×10-3 lx
    Fig. 10. Denoising results of parrot image under 2×10-3 lx
    Denoising results of butterfly image under 2×10-3 lx
    Fig. 11. Denoising results of butterfly image under 2×10-3 lx
    MethodsPSNR/dB(↑)SSIM(↑)
    10-1lx10-2lx10-3lxAverage10-1lx10-2lx10-3lxAverage
    Noisy images24.1720.3918.4120.990.543 90.409 20.355 00.436 0
    NLM26.9721.8219.5422.780.712 00.503 70.431 60.549 1
    DnCNN24.2620.4518.4521.050.549 50.413 30.357 80.440 2
    CBDnet24.3620.5018.5121.120.555 60.417 40.363 20.445 4
    VDN23.2121.2320.0021.480.565 10.449 00.408 00.474 0
    BM3D25.4426.6417.4523.180.865 70.878 60.800 10.848 1
    Ours33.2332.1132.8732.740.895 00.886 50.913 70.898 4
    Table 1. Comparison of objective evaluation indicators of different methods
    MethodsTime cost/sPlatform
    NLM564Python(CPU)
    DnCNN0.179Pytorch(GPU)
    CBDnet0.01Pytorch(GPU)
    VDN0.14Pytorch(GPU)
    BM3D1 868Python(CPU)
    Ours0.005Pytorch(GPU)
    Table 2. The running time of different methods for processing a single 1 920×1 080 image
    Xia WANG, Xin ZHANG, Gangcheng JIAO, Ye YANG, Hongchang CHENG, Bo YAN. Dual Residual Attention Network for ICMOS Sensing Image[J]. Acta Photonica Sinica, 2022, 51(6): 0610002
    Download Citation