• Laser & Optoelectronics Progress
  • Vol. 58, Issue 2, 0210024 (2021)
Cuijun Zhang1、2 and Na Zhao1、*
Author Affiliations
  • 1School of Information Engineering, Hebei GEO University, Shijiazhuang, Hebei 0 50031, China
  • 2Hebei Center for Ecological and Environmental Geology Research, Hebei GEO University, Shijiazhuang, Hebei 0 50031, China
  • show less
    DOI: 10.3788/LOP202158.0210024 Cite this Article Set citation alerts
    Cuijun Zhang, Na Zhao. Improved GrabCut Algorithm Based on Probabilistic Neural Network[J]. Laser & Optoelectronics Progress, 2021, 58(2): 0210024 Copy Citation Text show less
    Structure of image G
    Fig. 1. Structure of image G
    Structure diagram of the PNN
    Fig. 2. Structure diagram of the PNN
    Grayscale histograms of foreground and background. (a) Original image; (b) grayscale histogram of the foreground; (c) grayscale histogram of the background
    Fig. 3. Grayscale histograms of foreground and background. (a) Original image; (b) grayscale histogram of the foreground; (c) grayscale histogram of the background
    Flow chart of PNN_GrabCut algorithm
    Fig. 4. Flow chart of PNN_GrabCut algorithm
    Experimental example. (a) Original image; (b) label image
    Fig. 5. Experimental example. (a) Original image; (b) label image
    Segmentation results of different algorithms. (a) Original image; (b) GrabCut algorithm; (c) PNN_GrabCut algorithm; (d) algorithm of Ref. [12]; (d) algorithm of Ref. [16]
    Fig. 6. Segmentation results of different algorithms. (a) Original image; (b) GrabCut algorithm; (c) PNN_GrabCut algorithm; (d) algorithm of Ref. [12]; (d) algorithm of Ref. [16]
    Segmentation results of different algorithms. (a) Original image; (b) GrabCut algorithm; (c) PNN_GrabCut algorithm
    Fig. 7. Segmentation results of different algorithms. (a) Original image; (b) GrabCut algorithm; (c) PNN_GrabCut algorithm
    Pixel No.190-194195-199185-199180-18440-4445-49175-179155-159
    Amount141813721020584405354317272
    Table 1. Statistics of the high pixel value in foreground
    Pixel No.160-164165-169175-179155-159170-174180-184150-154145-149
    Amount3224230422532246217613301144414
    Table 2. Statistics of the high pixel value in background
    ClassTrain setValidation set
    Person5075526
    Plane13512
    Table 3. Experimental data
    σnP/%
    0.00051532.94
    0.00163112.11
    0.005241346.31
    0.0176614.70
    0.055099.77
    0.154010.37
    0.51983.80
    Table 4. PNN prediction results at different σ(experiment1)
    σnP/%
    0.0031917.92
    0.00437915.70
    0.005120649.98
    0.00643518.03
    0.0072028.37
    Table 5. PNN prediction results at different σ (experiment2)
    AlgorithmAverage value of F1Average timeF1 increase rate/%Time increase rate/%
    GrabCut0.8096.501//
    PNN_GrabCut0.8575.2295.9319.57
    Ref. [12]0.8274.7282.2227.27
    Ref. [16]0.8465.4194.5716.64
    Table 6. Average F1 and running time of different algorithms
    Cuijun Zhang, Na Zhao. Improved GrabCut Algorithm Based on Probabilistic Neural Network[J]. Laser & Optoelectronics Progress, 2021, 58(2): 0210024
    Download Citation