Laser & Optoelectronics Progress, Volume. 61, Issue 16, 1611012(2024)

Disparity-Guided Matching Metric Method for Light Field Features (Invited)

Meng Zhang1, Haiyan Jin1,2, Zhaolin Xiao1,2、*, and Fengyuan Zuo1
Author Affiliations
  • 1Department of Computer Science, Xi'an University of Technology, Xi'an 710048, Shaanxi, China
  • 2Shaanxi Key Laboratory for Network Computing and Security Technology, Xi'an 710048, Shaanxi, China
  • show less
    Figures & Tables(7)
    Overall framework of proposed light field feature matching
    Adaptive calibration module
    Feature matching results on light field images with simple geometric transformations (green lines represent correct matches, red lines represent incorrect matches)
    Feature matching results on light field images under challenging scenarios including illumination variations, non-Lambertian surfaces, and repetitive textures
    • Table 1. Quantitative comparison of light field feature matching with simple geometric transformations

      View table

      Table 1. Quantitative comparison of light field feature matching with simple geometric transformations

      MethodFDL-HCGHORB+GMS

      SuperPoint+

      SuperGlue

      SuperPoint+

      LightGlue

      FDL-Harris+

      proposed method

      sofa369/8/0.971485/5/0.99225/0/1.00754/0/1.00381/0/1.00
      tableware643/31/0.951620/2/0.99261/0/1.00464/0/1.00714/0/1.00
      decorations353/5/0.981402/35/0.97353/2/0.99645/9/0.98775/0/1.00
      wall corner13/8/0.62400/47/0.89231/24/0.90316/37/0.89244/24/0.91
    • Table 2. Comparison of light field image feature matching performance under challenging scenes involving illumination variations, non-Lambertian effects, and repetitive textures

      View table

      Table 2. Comparison of light field image feature matching performance under challenging scenes involving illumination variations, non-Lambertian effects, and repetitive textures

      MethodFDL-HCGHORB+GMS

      SuperPoint+

      SuperGlue

      SuperPoint+

      LightGlue

      FDL-Harris+

      proposed method

      Average precision0.760.900.850.920.96
      chess405/230/0.64679/2/0.99370/0/1.00714/5/0.99302/0/1.00
      office40/28/0.59261/41/0.86128/59/0.68336/32/0.91162/5/0.97
      toys58/31/0.65305/69/0.81193/30/0.86400/37/0.91298/9/0.97
      flower73/22/0.77452/69/0.86303/18/0.94586/28/0.95227/10/0.96
      games81/18/0.82396/30/0.92198/79/0.71258/47/0.84358/24/0.94
      kettle339/5/0.991471/10/0.99299/11/0.96516/18/0.96364/0/1.00
      toiletries126/25/0.83668/107/0.86168/38/0.81479/76/0.86420/44/0.91
    • Table 3. Comparative analysis of feature matching precision on light field matching dataset using various distance metrics

      View table

      Table 3. Comparative analysis of feature matching precision on light field matching dataset using various distance metrics

      MethodL1L2MahalanobisCosineCorrelationProposed method
      Average precision0.560.620.500.820.550.97
    Tools

    Get Citation

    Copy Citation Text

    Meng Zhang, Haiyan Jin, Zhaolin Xiao, Fengyuan Zuo. Disparity-Guided Matching Metric Method for Light Field Features (Invited)[J]. Laser & Optoelectronics Progress, 2024, 61(16): 1611012

    Download Citation

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

    Category: Imaging Systems

    Received: May. 16, 2024

    Accepted: Jun. 17, 2024

    Published Online: Aug. 12, 2024

    The Author Email: Zhaolin Xiao (xiaozhaolin@xaut.edu.cn)

    DOI:10.3788/LOP241287

    CSTR:32186.14.LOP241287

    Topics