• Laser & Optoelectronics Progress
  • Vol. 62, Issue 2, 0228001 (2025)
Jiale Fan*, Qiang Li, Ruifeng Zhang, and Xin Guan
Author Affiliations
  • School of Microelectronics, Tianjin University, Tianjin 300072, China
  • show less
    DOI: 10.3788/LOP241023 Cite this Article Set citation alerts
    Jiale Fan, Qiang Li, Ruifeng Zhang, Xin Guan. Spatial Spectral VAFormer Graph Convolution Hyperspectral Image Super-Resolution Network[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0228001 Copy Citation Text show less
    SSVF model schematic
    Fig. 1. SSVF model schematic
    Architecture of composite graph convolution module
    Fig. 2. Architecture of composite graph convolution module
    Schematic of multiscale mix convolution module
    Fig. 3. Schematic of multiscale mix convolution module
    Architecture of VAFormer
    Fig. 4. Architecture of VAFormer
    Spatial distribution diagram of pixels in CAVE、Harvard dataset
    Fig. 5. Spatial distribution diagram of pixels in CAVE、Harvard dataset
    Spectral information distribution of CAVE、Harvard dataset. (a) h1 image in Harvard dataset; (b) oil_painting image in CAVE dataset
    Fig. 6. Spectral information distribution of CAVE、Harvard dataset. (a) h1 image in Harvard dataset; (b) oil_painting image in CAVE dataset
    Heat map of correlation distribution between different spectral bands of CAVE、Harvard dataset
    Fig. 7. Heat map of correlation distribution between different spectral bands of CAVE、Harvard dataset
    Continuous density estimation between different spectral bands for CAVE dataset flower_ms
    Fig. 8. Continuous density estimation between different spectral bands for CAVE dataset flower_ms
    Heatmap of spatial correlation between different patches of CAVE、Harvard dataset
    Fig. 9. Heatmap of spatial correlation between different patches of CAVE、Harvard dataset
    Comparison of absolute error plots for local zoom in the 10th channel of the CAVE data cloth_ms. (a) Grand truth; (b) CSTF; (c) UAL; (d) TSFN; (e) PZnet; (f) FF-former; (g) LGAR; (h) SSVF
    Fig. 10. Comparison of absolute error plots for local zoom in the 10th channel of the CAVE data cloth_ms. (a) Grand truth; (b) CSTF; (c) UAL; (d) TSFN; (e) PZnet; (f) FF-former; (g) LGAR; (h) SSVF
    Comparison of absolute error plots for local zoom in the 10th channel of the Harvard data imga3. (a) Grand truth; (b) CSTF; (c) UAL; (d) TSFN; (e) PZnet; (f) FF-former; (g) LGAR; (h) SSVF
    Fig. 11. Comparison of absolute error plots for local zoom in the 10th channel of the Harvard data imga3. (a) Grand truth; (b) CSTF; (c) UAL; (d) TSFN; (e) PZnet; (f) FF-former; (g) LGAR; (h) SSVF
    Comparison of image element spectra at different positions of beads_ms for the CAVE dataset. (a) (160, 200); (b) (400, 400); (c) (200, 160)
    Fig. 12. Comparison of image element spectra at different positions of beads_ms for the CAVE dataset. (a) (160, 200); (b) (400, 400); (c) (200, 160)
    Comparison of image element spectra at different positions of imgh7 for the Harvard dataset. (a) (400, 500); (b) (400, 20); (c) (10, 10)
    Fig. 13. Comparison of image element spectra at different positions of imgh7 for the Harvard dataset. (a) (400, 500); (b) (400, 20); (c) (10, 10)
    IndicatorSpatial correlation threshold
    0.20.40.60.8
    PSNR51.24451.29251.29950.172
    SSIM0.99640.99640.99640.9958
    SAM1.7031.6811.6261.833
    MSE0.00280.00280.00280.0030
    Table 1. Effect of spatial correlation threshold on the model when the spectral correlation threshold is 0.2
    IndicatorSpectral correlation threshold
    0.20.40.60.8
    PSNR51.29951.20450.87149.106
    SSIM0.99640.99620.99590.9952
    SAM1.6261.7101.7661.997
    RMSE0.00280.00280.00300.0033
    Table 2. Effect of spectral correlation threshold on the model when the spatial correlation threshold is 0.6
    ModuleIndicator
    PANRSSIMSAMRMSE
    None47.86500.99412.10700.0039
    MMC50.38700.99521.92500.0031
    VAFormer50.17100.99501.83100.0031
    VAFormer+CGC50.94400.99561.70300.0029
    MMC+VAFormer+CGC51.29900.99641.62600.0028
    Table 3. Ablation studies by component
    FactorIndicatorCSTFUALTSFNPZnetFF-formerLGARSSVF
    ×4PSNR42.30043.96448.85350.28849.18650.97951.299
    SSIM0.96440.99200.99480.99560.99530.99600.9964
    SAM7.7163.0421.8691.9821.9971.8331.781
    RMSE0.00900.00780.00460.00340.00350.00280.0028
    ×8PSNR41.71742.58549.26649.65449.10849.73350.211
    SSIM0.96650.99100.99430.99450.99540.99470.9949
    SAM7.5023.8992.3112.1892.2692.1081.965
    RMSE0.00970.00900.00380.00370.00380.00370.0037
    ×16PSNR39.82440.10546.53346.99446.32246.91547.041
    SSIM0.96330.98860.99220.99200.99300.99220.9928
    SAM7.3294.2572.9212.6812.9612.8402.772
    RMSE0.01460.01130.00580.00570.00580.00570.0057
    Table 4. Results of different scale factors comparison algorithms on CAVE dataset
    FactorIndicatorCSTFUALTSFNPZnetFF-formerLGARSSVF
    ×4PSNR42.36244.85348.99349.38248.85649.41149.762
    SSIM0.97220.98650.98740.98730.98730.98740.9879
    SAM8.2574.7623.4883.2503.4223.3113.135
    RMSE0.00900.00690.00380.00360.00380.00360.0036
    ×8PSNR41.23343.36848.66148.86448.20348.83549.014
    SSIM0.97060.98370.98500.98490.98500.98500.9860
    SAM9.2145.0113.7633.9013.9553.8583.351
    RMSE0.00970.00860.00390.00380.00410.00380.0038
    ×16PSNR39.82242.56347.29947.23246.95847.23547.255
    SSIM0.95210.97880.98030.98010.98030.98030.9827
    SAM9.6215.8124.3444.2194.4254.2684.221
    RMSE0.01460.01130.00580.00570.00580.00570.0057
    Table 5. Results of different scale factors comparison algorithms on Harvard dataset
    Jiale Fan, Qiang Li, Ruifeng Zhang, Xin Guan. Spatial Spectral VAFormer Graph Convolution Hyperspectral Image Super-Resolution Network[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0228001
    Download Citation