• Laser & Optoelectronics Progress
  • Vol. 57, Issue 24, 241018 (2020)
Xiaosa Zhao1, Xijiang Chen1、*, Ya Ban2, Dandan Zhang1, and Lexian Xu1
Author Affiliations
  • 1School of Resource & Environment Engineering, Wuhan University of Technology, Wuhan, Hubei 430079, China;
  • 2Chongqing Institute of Metrology and Quality Inspection, Chongqing 401120, China
  • show less
    DOI: 10.3788/LOP57.241018 Cite this Article Set citation alerts
    Xiaosa Zhao, Xijiang Chen, Ya Ban, Dandan Zhang, Lexian Xu. Power Function-Weighted Image Stitching Method Involving Improved SURF and Cell Acceleration[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241018 Copy Citation Text show less
    Matching process of bidirectional consistency constraint
    Fig. 1. Matching process of bidirectional consistency constraint
    Fusion process of slow in and out weighted fusion algorithm
    Fig. 2. Fusion process of slow in and out weighted fusion algorithm
    h(t) function curves under different conditions
    Fig. 3. h(t) function curves under different conditions
    Power function weighted fusion process
    Fig. 4. Power function weighted fusion process
    Power function weighting coefficient of Cell acceleration
    Fig. 5. Power function weighting coefficient of Cell acceleration
    Matching results of image feature points under different luminance by different algorithms. (a) Traditional SURF algorithm; (b) proposed algorithm
    Fig. 6. Matching results of image feature points under different luminance by different algorithms. (a) Traditional SURF algorithm; (b) proposed algorithm
    Matching results of image feature points under different angles by different algorithms. (a) Traditional SURF algorithm; (b) proposed algorithm
    Fig. 7. Matching results of image feature points under different angles by different algorithms. (a) Traditional SURF algorithm; (b) proposed algorithm
    Matching results of image feature points under different resolutions by different algorithms. (a) Traditional SURF algorithm; (b) proposed algorithm
    Fig. 8. Matching results of image feature points under different resolutions by different algorithms. (a) Traditional SURF algorithm; (b) proposed algorithm
    Matching results of image feature points under different scales by different algorithms. (a) Traditional SURF algorithm; (b) proposed algorithm
    Fig. 9. Matching results of image feature points under different scales by different algorithms. (a) Traditional SURF algorithm; (b) proposed algorithm
    Two original image sequences with different brightness. (a) Image 1; (b) image 2
    Fig. 10. Two original image sequences with different brightness. (a) Image 1; (b) image 2
    Comparison of results of three fusion algorithms on images for different brightness. (a) Slow in and slow out weighted fusion algorithm; (b) Ref. [16]; (c) proposed algorithm
    Fig. 11. Comparison of results of three fusion algorithms on images for different brightness. (a) Slow in and slow out weighted fusion algorithm; (b) Ref. [16]; (c) proposed algorithm
    Two original image sequences with different angles. (a) Image 1; (b) image 2
    Fig. 12. Two original image sequences with different angles. (a) Image 1; (b) image 2
    Comparison of results of three fusion algorithms on images from different angles. (a) Slow in and slow out weighted fusion algorithm; (b) Ref. [16]; (c) proposed algorithm
    Fig. 13. Comparison of results of three fusion algorithms on images from different angles. (a) Slow in and slow out weighted fusion algorithm; (b) Ref. [16]; (c) proposed algorithm
    Two original image sequences. (a)Image 1 (b)Image 2
    Fig. 14. Two original image sequences. (a)Image 1 (b)Image 2
    Comparison of results of three fusion algorithms on images with different resolutions. (a) Gradual fade weighted fusion algorithm; (b) Ref. [16]; (c) proposed algorithm
    Fig. 15. Comparison of results of three fusion algorithms on images with different resolutions. (a) Gradual fade weighted fusion algorithm; (b) Ref. [16]; (c) proposed algorithm
    Two original image sequences of different heights. (a) Image 1; (b) image 2
    Fig. 16. Two original image sequences of different heights. (a) Image 1; (b) image 2
    Comparison of results of three fusion algorithms on images of different heights. (a) Slow in and slow out weighted fusion algorithm; (b) Ref. [16]; (c) proposed algorithm
    Fig. 17. Comparison of results of three fusion algorithms on images of different heights. (a) Slow in and slow out weighted fusion algorithm; (b) Ref. [16]; (c) proposed algorithm
    Time-consuming comparison histogram of 4 groups of experiments
    Fig. 18. Time-consuming comparison histogram of 4 groups of experiments
    MSE comparison of related algorithms
    Fig. 19. MSE comparison of related algorithms
    Information entropy data comparison
    Fig. 20. Information entropy data comparison
    Test imageRotate and zoomBrightnessBlurryPerspective
    Optimal threshold0.970.960.980.99
    Table 1. Test results of K values
    AlgorithmDifferent brightnessDifferent anglesDifferent resolutionsDifferent scales
    Correct matching rate /%Time /sCorrect matching rate /%Time/sCorrectmatching rate /%Time/sCorrectmatching rate /%Time/s
    Traditional SURF algorithm74.073.5965.523.5870.524.1678.253.49
    Proposed algorithm85.182.0176.802.0282.102.1090.152.00
    Table 2. Data comparison of related algorithms
    Xiaosa Zhao, Xijiang Chen, Ya Ban, Dandan Zhang, Lexian Xu. Power Function-Weighted Image Stitching Method Involving Improved SURF and Cell Acceleration[J]. Laser & Optoelectronics Progress, 2020, 57(24): 241018
    Download Citation