Chinese Journal of Lasers, Volume. 51, Issue 21, 2109001(2024)

Frequency Decomposition and Double-Branch Feature Extraction for Multispectral-Image-Compression Network

Dexiao Xu1、*, Fanqiang Kong1, Kun Wang2, Xu Fang3, and Murong Huang1
Author Affiliations
  • 1College of Astronautics, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, Jiangsu , China
  • 2The First Military Representative Office of Empty Equipment Stationed in Wuxi, Wuxi 214000, Jiangsu , China
  • 3AVIC Leihua Electronic Technology Institue, Wuxi 214000, Jiangsu , China
  • show less
    Figures & Tables(11)
    Principal network framework
    Frequency decomposition
    Frequency synthesis
    Feature extraction module
    Comparison results of each evaluation index in 8-band dataset. (a) PSNR; (b) MS-SSIM; (c) MSA
    Comparison results of each evaluation index in 12-band dataset. (a) PSNR; (b) MS-SSIM; (c) MSA
    Detail comparison of different methods for 8-band dataset. (a) Original image; (b) proposed method; (c) DBSSFE; (d) VCC; (e) JPEG2000
    Detail comparison of different methods for 12-band dataset. (a) Original image; (b) proposed method; (c) DBSSFE; (d) VCC; (e) JPEG2000
    • Table 1. Test data for each method in 8-band dataset at 0.223, 0.304, and 0.429 three bit rates

      View table

      Table 1. Test data for each method in 8-band dataset at 0.223, 0.304, and 0.429 three bit rates

      MethodPSNR /dBMS-SSIM /dBMSA /rad
      JPEG200031.32032.17033.93015.77017.85019.3900.057890.051590.04709
      3D-SPHIT33.81035.09037.15014.01015.42017.1300.071650.061880.05488
      VCC38.03040.34042.53020.53022.65024.3200.042030.032950.02663
      Joint40.67041.53042.02024.29025.12025.3000.029920.029440.02827
      DBSSFE39.24041.39042.45023.03023.84024.9000.031250.030500.28730
      Proposed41.03042.68043.65024.67025.82026.7600.028870.026410.02504
    • Table 2. Test data for each method in 12-band dataset at 0.223, 0.304, and 0.429 three bit rates

      View table

      Table 2. Test data for each method in 12-band dataset at 0.223, 0.304, and 0.429 three bit rates

      MethodPSNR /dBMS-SSIM /dBMSA /rad
      JPEG200052.12055.85056.45026.37031.27032.9500.080760.056450.05176
      3D-SPHIT53.22056.54057.29028.04032.40034.2200.070010.050450.04206
      VCC57.02059.87061.05032.61036.17037.6100.052530.040060.03571
      Joint60.05061.89062.74036.02038.52039.3600.039830.029630.02679
      DBSSFE58.15061.00061.25035.05037.11037.5300.050860.039530.03868
      Proposed60.68063.58064.26036.59039.94040.3400.035270.025610.02501
    • Table 3. Encoding and decoding time test data of each method

      View table

      Table 3. Encoding and decoding time test data of each method

      MethodEncode time /msDecode time /msMethodEncode time /msDecode time /ms
      CPUGPUCPUGPU
      JPEG2000247.3187.7Joint1961.35780.8
      3D-SPHIT1481.4758.1DBSSFE2391.81644.52236.7196.3
      VVC1816.2907.8Proposed2548.3133.410559.7108.3
    Tools

    Get Citation

    Copy Citation Text

    Dexiao Xu, Fanqiang Kong, Kun Wang, Xu Fang, Murong Huang. Frequency Decomposition and Double-Branch Feature Extraction for Multispectral-Image-Compression Network[J]. Chinese Journal of Lasers, 2024, 51(21): 2109001

    Download Citation

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

    Category: holography and information processing

    Received: Apr. 1, 2024

    Accepted: Jun. 11, 2024

    Published Online: Oct. 29, 2024

    The Author Email:

    DOI:10.3788/CJL240727

    CSTR:32183.14.CJL240727

    Topics