Journal of South China University of Technology (Natural Science Edition) ›› 2015, Vol. 43 ›› Issue (9): 20-26,33.doi: 10.3969/j.issn.1000-565X.2015.09.004

• Electronics, Communication & Automation Technology • Previous Articles     Next Articles

A Dynamic Gradient Vector Flow Model

Zhou Zhi-heng  Zhong Hui-qiang  Dai Ming   

  1. School of Electronic and Information Engineering，South China University of Technology，Guangzhou 510640，Guangdong，China
  • Received:2015-01-01 Revised:2015-04-23 Online:2015-09-25 Published:2015-09-07
  • Contact: 周智恒(1977-)，男，博士，教授，主要从事图像分析和自适应信号处理研究． E-mail: zhouzh@scut.edu.cn
  • About author:周智恒(1977-)，男，博士，教授，主要从事图像分析和自适应信号处理研究．
  • Supported by:
     Supported by the National Natural Science Foundation of China(61372142)，the Joint Fund of the National Natural Science Foundation of China and Guangdong Province(U1401252)and the Science and Technology Planning Projects of Guang dong Province(2013A011403003，2013B010102004)

PDF

84

Abstract: Traditional active contour models based on the gradient vector flow can only produce static force field，in
which the equilibrium problem often occurs and it causes a difficulty in the convergence of the contour curve to a long concave boundary． In order to solve this problem，a dynamic gradient vector flow model is proposed in this paper． In the model，first，a dynamic force field is generated by adopting an indicative function relevant to the evolving contour curve to weigh the edge gradient map． Then，the edge stopping function is employed to control the convergence of the evolving contour curve． The proposed model makes full use of the information of the evolving contour curve，and thus it avoids the premature convergence caused by the equilibrium problem of static external force field and pushes the contour to evolve to the concavity boundary． Simulation results show that，in comparison with the traditional models，the proposed model can segment the long concave boundary of the object successfully and achieves better segmentation results in extracting the complex boundary of the object．

Key words: image segmentation, active contour model, dynamic gradient vector flow

CLC Number: