Chinese Journal of Liquid Crystals and Displays, Volume. 38, Issue 5, 617(2023)

Real-time automatic image cropping based on portrait detection

Yu-hang WU1, Shan-ling LIN1,3、*, Zhi-xian LIN1,2,3, and Tai-liang GUO2,3
Author Affiliations
  • 1School of Advanced Manufacturing,Fuzhou University,Quanzhou 362200,China
  • 2School of Physics and Information Engineering,Fuzhou University,Fuzhou 350116,China
  • 3China Fujian Photoelectric Information Science and Technology Innovation Laboratory,Fuzhou 350116,China
  • show less
    Figures & Tables(11)
    Principle of portrait composition
    Cropping algorithm flow
    Detection network architecture
    Flow chart of intelligent composition
    Multiple portrait cropping result
    Single portrait cropping result 1
    Single portrait cropping result 2
    Results comparion of the method in this paper with the existing automatic image cropping methods
    • Table 1. Intelligent composition-solving the main character common algorithm pseudocode

      View table
      View in Article

      Table 1. Intelligent composition-solving the main character common algorithm pseudocode

      算法I:主要人物共同中心点坐标求解算法

      输入:bounding box坐标(x1iy1ix2iy2i),i=1,2…n

      输出:主要人物的共同中心点坐标center_main(XY);

      begin

      //遍历bounding box

      for i∈{0,…n}do;

      Si=(x2i-x1i)*(y2i-y1i

      xi=(x1i+x2i)/2;

      yi=(y1i+y2i)/2;

      //求出面积最大的bounding box并加入group

      Smax=max{S1,S2,…. Sn};

      group.appendmax);

      X=xmax;Y=ymax

      //加入低于阈值的索引i 到group

      for i∈{0,…n}do

      if(X-xi<Thand(Y-yi<Th

      group.append(i);

      //求group中所以索引值对应的bounding box

      的中心点坐标均值

      X=mean(xgroup[i);

      Y=mean(ygroup[i);

      end

    • Table 2. Comparison of the network performance before and after the transformation

      View table
      View in Article

      Table 2. Comparison of the network performance before and after the transformation

      MethodInputParams/MFLOPs/GAP
      Centernet-resnet1851217.4639.8677.19
      Ours5122.312.1680.53
    • Table 3. Comparison of the FPS with existing automated image cropping methods

      View table
      View in Article

      Table 3. Comparison of the FPS with existing automated image cropping methods

      MethodFPS(GPU)FPS(CPU)
      A2-RL50.05
      VPN750.8
      VFN0.50.005
      VEN0.20.002
      Ours775.02
    Tools

    Get Citation

    Copy Citation Text

    Yu-hang WU, Shan-ling LIN, Zhi-xian LIN, Tai-liang GUO. Real-time automatic image cropping based on portrait detection[J]. Chinese Journal of Liquid Crystals and Displays, 2023, 38(5): 617

    Download Citation

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

    Category: Research Articles

    Received: Aug. 1, 2022

    Accepted: --

    Published Online: Jul. 4, 2023

    The Author Email: Shan-ling LIN (sllin@fzu.edu.cn)

    DOI:10.37188/CJLCD.2022-0258

    Topics