• Laser & Optoelectronics Progress
  • Vol. 58, Issue 2, 0215008 (2021)
Hong Xiao, Chuan Tian*, Yi Zhang, Bo Wei, and Jiaqi Kang
Author Affiliations
  • School of Advanced Manufacturing Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
  • show less
    DOI: 10.3788/LOP202158.0215008 Cite this Article Set citation alerts
    Hong Xiao, Chuan Tian, Yi Zhang, Bo Wei, Jiaqi Kang. Stereo Matching Algorithm Based on Improved Census Transform and Gradient Fusion[J]. Laser & Optoelectronics Progress, 2021, 58(2): 0215008 Copy Citation Text show less
    Flow chart of our algorithm
    Fig. 1. Flow chart of our algorithm
    Census transform windows obtained by different algorithms. (a) Different windows; (b) center pixel value; (c) average value of window pixels; (d) our algorithm
    Fig. 2. Census transform windows obtained by different algorithms. (a) Different windows; (b) center pixel value; (c) average value of window pixels; (d) our algorithm
    Disparity images obtained by different algorithms. (a) Left image; (b) right image; (c) true disparity image; (d) CT; (e) impro-CT; (f) GRD; (g) our algorithm
    Fig. 3. Disparity images obtained by different algorithms. (a) Left image; (b) right image; (c) true disparity image; (d) CT; (e) impro-CT; (f) GRD; (g) our algorithm
    Aloe disparity images obtained by our algorithm. (a) Left image; (b) right image; (c) real disparity image; (d) disparity image in multi-scale space; (e) final disparity image
    Fig. 4. Aloe disparity images obtained by our algorithm. (a) Left image; (b) right image; (c) real disparity image; (d) disparity image in multi-scale space; (e) final disparity image
    Disparity images obtained by different algorithms under different conditions. (a) Left image; (b) right image; (c) ground truth; (d) CT; (e) Impro-CT; (f) GRD; (g) our algorithm
    Fig. 5. Disparity images obtained by different algorithms under different conditions. (a) Left image; (b) right image; (c) ground truth; (d) CT; (e) Impro-CT; (f) GRD; (g) our algorithm
    Disparity images of the actual scene. (a) Left image; (b) right image; (c) disparity images generated by our algorithm
    Fig. 6. Disparity images of the actual scene. (a) Left image; (b) right image; (c) disparity images generated by our algorithm
    ParameterωkTdsλ
    Value9159
    Table 1. Parameter of our algorithm
    AlgorithmBaby1Wood1AloeMonopolyPlasticLampshade2Average
    CT1.928.391.904.893.275.994.39
    Impro-CT1.877.392.814.774.045.235.23
    GRD3.597.662.177.326.099.715.37
    Ours1.858.031.754.723.675.014.17
    Table 2. Mismatch rates of different cost algorithms (non-occluded area) unit: %
    AlgorithmBaby1Wood1AloeMonopolyPlasticLampshade2Average
    CT2.3812.192.199.0410.277.327.23
    Impro-CT2.5410.602.098.519.537.306.76
    GRD4.4711.582.7811.688.9213.778.87
    Ours2.3012.101.978.296.095.496.04
    Table 3. Mismatch rates of different cost algorithms (occlusion area) unit: %
    Hong Xiao, Chuan Tian, Yi Zhang, Bo Wei, Jiaqi Kang. Stereo Matching Algorithm Based on Improved Census Transform and Gradient Fusion[J]. Laser & Optoelectronics Progress, 2021, 58(2): 0215008
    Download Citation