Laser & Optoelectronics Progress, Volume. 62, Issue 2, 0237009(2025)

HDR Image Reconstruction Algorithm Based on Masked Transformer

Zuheng Zhang*, Xiaodong Chen, Yi Wang, and Huaiyu Cai
Author Affiliations
  • Key Laboratory of Opto-Electronics Information Technology of Ministry of Education, College of Precision Instrument and Opto-Electronic Engineering, Tianjin University, Tianjin 300072, China
  • show less
    Figures & Tables(13)
    Algorithmic framework for HDR image reconstruction based on bracketed image stack generation
    Schematic of bracketed image stack generation using a recursive approach
    Masked-transformer-based image generation network and related modules
    Mask acquisition method
    Role of global and local information in bracketed image stack generation
    Masking operation realization process
    Comparison of different algorithms for HDR image reconstruction. (a) Input image; (b) ground truth; (c) Ref. [15]; (d) Ref. [13]; (e) Ref. [20]; (f) proposed method
    Residual blocks for replacing MT modules
    Subjective evaluation of the results of ablation experiments. (a) Input image; (b) ground truth; (c) group 1; (d) group 2; (e) group 3
    Sample of attention map. (a) First layer; (b) bottleneck; (c) final layer
    • Table 1. Quantitative comparison of bracketed image stack generation

      View table

      Table 1. Quantitative comparison of bracketed image stack generation

      ImageMethodPSNR /dBSSIMMS-SSIM
      mσmσmσ
      EV+3Proposed30.953.680.9350.0590.9820.014
      Ref. [2028.884.900.8780.0560.9620.028
      EV+2Proposed31.453.400.9390.0560.9860.012
      Ref. [2030.543.030.9300.0300.9770.013
      EV+1Proposed33.743.730.9570.0360.9910.008
      Ref. [2030.753.250.9360.0420.9850.009
      EV-1Proposed30.973.690.9300.0430.9890.009
      Ref. [2030.253.350.9160.0440.9790.010
      EV-2Proposed30.424.830.9270.0490.9900.017
      Ref. [2028.303.030.9060.0550.9680.014
      EV-3Proposed28.463.520.8900.0600.9630.021
      Ref. [2027.623.630.8590.0570.9600.017
    • Table 2. Quantitative comparison of HDR image generation

      View table

      Table 2. Quantitative comparison of HDR image generation

      ImageMethodPSNR /dB (Reinhard’s TMO)PSNR /dB (Kim&Kautz’s TMO)HDR-VDP-2 (Q score)
      mσmσmσ
      VDSProposed32.143.4030.323.7461.095.95
      Ref. [1527.721.9826.002.1655.646.24
      Ref. [1318.261.9317.333.2142.494.62
      Ref. [2029.772.9828.072.8355.745.35

      HDR

      Eye

      Proposed23.412.0423.312.4946.114.61
      Ref. [1522.813.2622.562.5645.845.45
      Ref. [1316.652.0017.313.0943.164.61
      Ref. [2022.082.4622.732.6645.954.59
    • Table 3. Quantitative comparison of HDR image generation

      View table

      Table 3. Quantitative comparison of HDR image generation

      ImageGroupPSNR /dB (Reinhard’s TMO)PSNR /dB (Kim&Kautz’s TMO)HDR-VDP-2 (Q score)
      mσmσmσ
      VDSGroup 129.834.5328.063.6154.035.12
      Group 231.353.9329.054.5060.085.61
      Group 332.143.4030.323.7461.095.95

      HDR

      eye

      Group 121.242.1320.572.8043.894.78
      Group 223.232.3422.882.0245.944.60
      Group 323.412.0423.312.4946.114.61
    Tools

    Get Citation

    Copy Citation Text

    Zuheng Zhang, Xiaodong Chen, Yi Wang, Huaiyu Cai. HDR Image Reconstruction Algorithm Based on Masked Transformer[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0237009

    Download Citation

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

    Category: Digital Image Processing

    Received: Apr. 22, 2024

    Accepted: May. 31, 2024

    Published Online: Jan. 3, 2025

    The Author Email:

    DOI:10.3788/LOP241132

    CSTR:32186.14.LOP241132

    Topics