• Laser & Optoelectronics Progress
  • Vol. 59, Issue 8, 0811008 (2022)
Weihua Liu and Biyan Ma*
Author Affiliations
  • School of Communication and Information Engineering, Xi'an University of Posts & Telecommunications, Xi'an , Shaanxi 710121, China
  • show less
    DOI: 10.3788/LOP202259.0811008 Cite this Article Set citation alerts
    Weihua Liu, Biyan Ma. Multiexposure Image Fusion Method Based on Feature Weight of Image Sequence[J]. Laser & Optoelectronics Progress, 2022, 59(8): 0811008 Copy Citation Text show less
    Flow chart of multiexposure fusion method based on full sequence feature weight of image
    Fig. 1. Flow chart of multiexposure fusion method based on full sequence feature weight of image
    Comparison of local brightness weight between well exposed image and overexposed image. (a) Input images; (b) local brightness weight of rectangular box area
    Fig. 2. Comparison of local brightness weight between well exposed image and overexposed image. (a) Input images; (b) local brightness weight of rectangular box area
    Weight comparison
    Fig. 3. Weight comparison
    Process of obtaining global brightness weight. (a) Luminance maps; (b) average brightness maps of image sequence; (c) distance maps; (d) global brightness weights
    Fig. 4. Process of obtaining global brightness weight. (a) Luminance maps; (b) average brightness maps of image sequence; (c) distance maps; (d) global brightness weights
    Comparison between Canny local gradient and proposed gradient weight. (a) Canny local gradient; (b) proposed gradient weight
    Fig. 5. Comparison between Canny local gradient and proposed gradient weight. (a) Canny local gradient; (b) proposed gradient weight
    Results of Tower multiexposure image sequence. (a) Input image sequence; (b) local brightness weight; (c) global brightness weight; (d) gradient weight; (e) final weight; (f) fusion image
    Fig. 6. Results of Tower multiexposure image sequence. (a) Input image sequence; (b) local brightness weight; (c) global brightness weight; (d) gradient weight; (e) final weight; (f) fusion image
    Results of Venice multiexposure image sequence. (a) Input image sequence; (b) local brightness weight; (c) global brightness weight; (d) gradient weight; (e) final weight; (f) fusion image
    Fig. 7. Results of Venice multiexposure image sequence. (a) Input image sequence; (b) local brightness weight; (c) global brightness weight; (d) gradient weight; (e) final weight; (f) fusion image
    Fusion results of different algorithms. (a)‒(e) Algorithms in references [16], [14], [18], [12], [11]; (f) proposed algorithm
    Fig. 8. Fusion results of different algorithms. (a)‒(e) Algorithms in references [16], [14], [18], [12], [11]; (f) proposed algorithm
    Local enlarged images of fusion result. (a)‒(e) Algorithms in references [16], [14], [18], [12], [11]; (f) proposed algorithm
    Fig. 9. Local enlarged images of fusion result. (a)‒(e) Algorithms in references [16], [14], [18], [12], [11]; (f) proposed algorithm
    Influence of weight on cave image quality.(a) Fusion image based on W1; (b) fusion image based on W1W2; (c) fusion image based on W1W2W3
    Fig. 10. Influence of weight on cave image quality.(a) Fusion image based on W1; (b) fusion image based on W1W2; (c) fusion image based on W1W2W3

    Image

    sequence

    26

    2012

    27

    2011

    20

    2012

    21

    2013

    16

    2009

    14

    2015

    18

    2018

    12

    2020

    11

    2020

    Proposed algorithm
    Average0.910.9310.9450.9650.9770.9740.9780.9430.8820.980
    Balloons0.9130.9180.9410.9480.9630.9690.9710.9120.8990.972
    Cave0.9340.9180.9230.9780.9800.9750.9770.9460.9320.982
    Chinese garden0.9270.9670.9510.9840.9880.9890.9900.9900.8610.99
    Farmhouse0.9320.9470.9590.9850.9830.9810.9780.9830.930.983
    Lamp0.8710.8290.9330.9340.9450.940.9540.8240.8860.955
    Landscape0.9410.9440.9480.9420.9910.9760.9810.9910.8350.984
    Madison0.8640.9450.9490.9680.9740.9770.9780.9490.9290.982
    Office0.9000.9610.9540.9670.9860.9850.9890.9820.8860.991
    Tower0.9320.9390.9500.9860.9810.9860.9870.9880.9000.985
    Venice0.8890.9420.9370.9540.9780.9660.9730.8640.7670.976
    Table 1. MEF-SSIM scores of different algorithms
    Image sequence

    16

    2009

    14

    2015

    18

    2018

    12

    2020

    11

    2020

    Proposed algorithm
    Average7.4827.5887.6497.6227.6347.652
    Balloons7.5967.5177.8407.6357.3887.838
    Cave7.1017.5807.5797.6647.3777.599
    Chinese garden7.8257.7777.7517.7457.8557.776
    Farmhouse7.3177.2407.3977.4517.5357.358
    Lamp7.3727.3457.7247.7877.7287.758
    Landscape7.3957.6347.4427.3417.4897.426
    Madison7.7827.7507.7937.8387.7307.779
    Office7.3207.5467.5267.4637.6777.525
    Tower7.5827.5737.6347.6447.7837.686
    Venice7.5317.9167.8057.6507.7757.776
    Table 2. Average information entropy of fusion results of different algorithms
    Image sequenceW1W2W3W1W2W1W2W3
    Average0.9730.8740.9530.9750.980
    Balloons0.9700.9640.9350.9710.972
    Cave0.9510.6740.9520.9620.982
    Chinese garden0.9880.9150.9650.9870.990
    Farmhouse0.9800.8770.9690.9810.983
    Lamp0.9510.8660.9310.9510.955
    Landscape0.9830.9560.9610.9830.984
    Madison0.9810.7760.9430.9820.982
    Office0.9890.9170.9470.9900.991
    Tower0.9680.8940.9730.9710.985
    Venice0.9700.9000.9510.9710.976
    Table 3. MEF-SSIM scores with different weight combinations
    Image sequence (NTime
    Average1.34
    Balloons (9)1.45
    Cave (4)0.82
    Chinese garden (3)0.69
    Farmhouse (3)0.67
    Lamp (6)1.01
    Landscape (3)0.62
    Madison (30)5.57
    Office (6)1.04
    Tower (3)0.85
    Venice (3)0.65
    Table 4. Execution time of proposed method
    Weihua Liu, Biyan Ma. Multiexposure Image Fusion Method Based on Feature Weight of Image Sequence[J]. Laser & Optoelectronics Progress, 2022, 59(8): 0811008
    Download Citation