• Laser & Optoelectronics Progress
  • Vol. 55, Issue 12, 121003 (2018)
Heng Zhao* and Weisheng An**
Author Affiliations
  • School of Mechanical Engineering, Southwest Jiaotong University, Chengdu, Sichuan 610031, China
  • show less
    DOI: 10.3788/LOP55.121003 Cite this Article Set citation alerts
    Heng Zhao, Weisheng An. Image Salient Object Detection Combined with Deep Learning[J]. Laser & Optoelectronics Progress, 2018, 55(12): 121003 Copy Citation Text show less
    Schematic of algorithm
    Fig. 1. Schematic of algorithm
    Examples of image processing. (a) Initial image; (b) convolved images; (c) clustered images; (d) CRF optimized images; (e) truth value chart
    Fig. 2. Examples of image processing. (a) Initial image; (b) convolved images; (c) clustered images; (d) CRF optimized images; (e) truth value chart
    Loss L versus number of iterations
    Fig. 3. Loss L versus number of iterations
    Comparison among P-R curves by several algorithms. (a) ECSSD; (b) DUT-ORMON
    Fig. 4. Comparison among P-R curves by several algorithms. (a) ECSSD; (b) DUT-ORMON
    Comparison among image salient detection results by several algorithms. (a) Initial images; (b) GT; (c) proposed algorithm; (d) DCL; (e) DRFI; (f) DS; (g) ELD; (h) FT; (i) MDF; (j) QCUT; (k) RC; (l) RFCN
    Fig. 5. Comparison among image salient detection results by several algorithms. (a) Initial images; (b) GT; (c) proposed algorithm; (d) DCL; (e) DRFI; (f) DS; (g) ELD; (h) FT; (i) MDF; (j) QCUT; (k) RC; (l) RFCN
    AlgorithmECSSDDUT-OMRON
    FMAEFMAE
    Proposed0.8770.0460.6890.048
    DCL0.8630.0470.6810.053
    DRFI0.7320.0800.5450.083
    DS0.8240.0750.5980.062
    ELD0.8140.0500.6090.059
    FT0.3790.1210.3110.106
    MDF0.7950.0790.6030.064
    QCUT0.7430.1210.4950.072
    RC0.6800.0980.6780.110
    RFCN0.8730.0470.6810.057
    Table 1. Comparison of F and MAE
    AlgorithmProposedDCLDSELDMDFRFCN
    Time /s0.31.51.10.584.6
    Table 2. Comparison of time consuming
    Heng Zhao, Weisheng An. Image Salient Object Detection Combined with Deep Learning[J]. Laser & Optoelectronics Progress, 2018, 55(12): 121003
    Download Citation