Chinese Journal of Liquid Crystals and Displays, Volume. 40, Issue 7, 1056(2025)

Image super-resolution reconstruction based on multidimensional attention network

Xing HE1,2, Lei WANG1,2、*, Pengchao ZHANG1,2, Shusheng WANG1,2, and Heng ZHANG1,2
Author Affiliations
  • 1School of Mechanical Engineering,Shaanxi University of Technology,Hanzhong 723001,China
  • 2Shaanxi Key Laboratory of Industrial Automation,Hanzhong 723001,China
  • show less
    Figures & Tables(11)
    Structural diagram of the SRDiff network
    Structural diagram of the LR encoder of the conditional noise predictor in this paper
    MAT general structure and CAM structure diagram
    Reconstruction effect of “woman” in Set5
    Reconstruction effect of “ppt3” in Set14
    Reconstruction effect of “img082” in BSD100
    Reconstruction effect of “image001” in Urban100
    • Table 1. PSNR values of different models

      View table
      View in Article

      Table 1. PSNR values of different models

      DatasetScaleBicubicVDSRSRGANSRDiffSRFlowIDMOurs
      Set5233.7637.5334.7735.4637.7437.8538.01
      Set1430.2333.0231.5733.8832.8933.4633.69
      BSD10029.5931.9131.6932.2431.8032.0532.26
      Urban10026.8630.7633.0232.9632.5331.7632.98
      DIV2K10032.5434.2233.6334.3633.6934.4234.46
      Set5428.3931.3529.4829.5930.0229.7630.25
      Set1426.0928.0825.9227.7828.5328.6528.72
      BSD10025.9727.2825.2427.3927.0527.2427.53
      Urban10023.1525.2925.4425.8926.2926.0626.42
      DIV2K10026.8327.1527.2827.1427.0927.2927.34
    • Table 2. SSIM values of different models

      View table
      View in Article

      Table 2. SSIM values of different models

      DatasetScaleBicubicVDSRSRGANSRDiffSRFlowIDMOurs
      Set520.9240.9580.9390.9430.9350.9480.958
      Set140.8670.9120.8870.9040.9160.9080.917
      BSD1000.8460.8940.8570.8950.8930.8890.898
      Urban1000.8410.9130.8620.9230.9180.9250.927
      DIV2K1000.9060.8760.9120.9150.9090.9140.924
      Set540.8150.8810.8430.8760.8820.8760.895
      Set140.7020.7640.7370.7430.7400.7540.765
      BSD1000.6680.7280.6630.7240.7350.7290.738
      Urban1000.6560.7560.7350.7750.7680.7820.783
      DIV2K1000.7740.8030.7340.7650.7830.7860.821
    • Table 3. Ablation experiments for MAT effectiveness

      View table
      View in Article

      Table 3. Ablation experiments for MAT effectiveness

      Exp.CABW-MSASAMSet5Set14
      PSNR/SSIMPSNR/SSIM
      1××29.87/0.87228.66/0.754
      2××29.96/0.88628.56/0.742
      3××30.01/0.88928.71/0.756
      4×30.20/0.89228.69/0.761
      5×30.02/0.88828.70/0.763
      630.25/0.89528.72/0.765
    • Table 4. Ablation experiments for RPDB effectiveness

      View table
      View in Article

      Table 4. Ablation experiments for RPDB effectiveness

      方法Params/kFLOPs/GPSNR/dBSSIM
      RRDB107 226.0527.120.795
      RRDB+PConv105 326.0727.290.816
      RRDB+PConv+MAT104 826.0627.340.821
    Tools

    Get Citation

    Copy Citation Text

    Xing HE, Lei WANG, Pengchao ZHANG, Shusheng WANG, Heng ZHANG. Image super-resolution reconstruction based on multidimensional attention network[J]. Chinese Journal of Liquid Crystals and Displays, 2025, 40(7): 1056

    Download Citation

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

    Category:

    Received: Mar. 18, 2025

    Accepted: --

    Published Online: Aug. 11, 2025

    The Author Email: Lei WANG (leiwang@xaut.edu.cn)

    DOI:10.37188/CJLCD.2025-0058

    Topics