• Laser & Optoelectronics Progress
  • Vol. 57, Issue 8, 081504 (2020)
Xinchun Li1, Xinyong Yin1、*, and Sen Lin1、2、3
Author Affiliations
  • 1School of Electronic and Information Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China
  • 2State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, Liaoning 110016, China
  • 3Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang, Liaoning 110016, China
  • show less
    DOI: 10.3788/LOP57.081504 Cite this Article Set citation alerts
    Xinchun Li, Xinyong Yin, Sen Lin. Stereo Matching by Improved Window Characteristics and Differential Operators[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081504 Copy Citation Text show less
    Direct relevant neighborhood of matching points
    Fig. 1. Direct relevant neighborhood of matching points
    Window shape and weight allocation
    Fig. 2. Window shape and weight allocation
    Main flow chart of matching in this paper
    Fig. 3. Main flow chart of matching in this paper
    Sample datasets diagrams (Jadeplant). (a) Left image; (b) right image; (c) ground truth; (d) mask
    Fig. 4. Sample datasets diagrams (Jadeplant). (a) Left image; (b) right image; (c) ground truth; (d) mask
    Mismatching rate of different window shapes
    Fig. 5. Mismatching rate of different window shapes
    Different types of matching windows
    Fig. 6. Different types of matching windows
    Example of the effect improvement of proposed algorithm and the original method. (a) Left images; (b) ground truth images; (c) ELAS; (d) proposed algorithm
    Fig. 7. Example of the effect improvement of proposed algorithm and the original method. (a) Left images; (b) ground truth images; (c) ELAS; (d) proposed algorithm
    bad index comparison
    Fig. 8. bad index comparison
    avgErr index comparison chart
    Fig. 9. avgErr index comparison chart
    Sample graphs of effectiveness improvement. (a) Left images; (b) ground truth; (c) GA-Net; (d) GwcNet; (e) Census; (f) SGM; (g) proposed algorithm
    Fig. 10. Sample graphs of effectiveness improvement. (a) Left images; (b) ground truth; (c) GA-Net; (d) GwcNet; (e) Census; (f) SGM; (g) proposed algorithm
    Self-collecting images. (a) Left images; (b) GA-Net; (c) GwcNet; (d) Census; (e) SGM; (f) proposed algorithm
    Fig. 11. Self-collecting images. (a) Left images; (b) GA-Net; (c) GwcNet; (d) Census; (e) SGM; (f) proposed algorithm
    Data nameEbad /%Eavg /%Data name
    ELASProposedELASProposed
    Adirondack12.07011.8900.8800.877Adirondack
    ArtL19.80019.2701.7601.630ArtL
    Jadeplant34.68034.6601.1701.170MotorcycleE
    MotorcycleE13.45013.2102.2102.080PianoL
    PianoL37.52036.5301.9601.930Pipes
    Pipes15.30014.7601.9001.800layroom
    Playroom26.68026.0404.5504.270Playtable
    Playtable26.38026.2900.9300.920PlaytableP
    PlaytableP14.84014.7400.8900.890Recycle
    Recycle17.85016.8500.9100.910Teddy
    Vintage33.69033.6602.4302.430Vintage
    Table 1. Performance comparison of Gaussian weight window under two indices
    Data nameEbad /%Eavg /%Data name
    ELASProposedELASProposed
    ArtL19.8019.730.880.86Adirondack
    Jadeplant34.6829.198.856.33Jadeplant
    MotorcycleE13.4513.091.171.13MotorcycleE
    Piano22.0421.922.212.22PianoL
    PianoL37.5237.501.961.97Pipes
    Pipes15.3015.190.930.92PlaytableP
    Playroom26.6825.630.910.89Teddy
    PlaytableP14.8414.79---
    Recycle17.8517.24---
    Table 2. Performance comparison of Scharr filter under two indices
    Data nameEbad /%Eavg /%Data name
    ELASProposedELASProposed
    Adirondack12.0711.810.880.83Adirondack
    ArtL19.8019.551.761.74ArtL
    Jadeplant34.6827.858.955.99Jadeplant
    MotorcycleE13.4513.301.171.15Motorcycle
    Piano22.0421.851.171.16MotorcycleE
    PianoL37.5236.032.211.94PianoL
    Pipes15.3015.301.961.94Pipes
    Playroom26.6825.630.930.92PlaytableP
    Recycle17.8517.121.971.92Shelves
    Teddy11.4210.880.910.83Teddy
    Vintage33.6932.332.432.39Vintage
    Table 3. Performance comparison of proposed algorithm under two indices
    Xinchun Li, Xinyong Yin, Sen Lin. Stereo Matching by Improved Window Characteristics and Differential Operators[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081504
    Download Citation