Computer Engineering and Applications ›› 2018, Vol. 54 ›› Issue (20): 213-218.DOI: 10.3778/j.issn.1002-8331.1707-0057

Previous Articles     Next Articles

Double signed pressure force function-based active contour model for image segmentation

SUN Lin1, KE Zhengyou2, FENG Xiaobo3, XU Jiucheng1, XUE Zhan’ao1   

  1. 1.College of Computer & Information Engineering, Henan Normal University, Xinxiang, Henan 453007, China
    2.School of Information Science & Engineering, Lanzhou University, Lanzhou 730000, China
    3.College of Information Science & Engineering, Hunan University, Changsha 410082, China
  • Online:2018-10-15 Published:2018-10-19

基于双符号压力函数的活动轮廓图像分割方法

孙  林1,柯正友2,冯小博3,徐久成1,薛占熬1   

  1. 1.河南师范大学 计算机与信息工程学院,河南 新乡 453007
    2.兰州大学 信息科学与工程学院,兰州 730000
    3.湖南大学 信息科学与工程学院,长沙 410082

Abstract: In order to solve the problems that geodesic active contour model and Chan-Vese(CV) model cannot segment the weak boundary and the gray uneven images at the same time, this paper presents an improved active contour model for image segmentation based on Double Signed Pressure Force functions(DSPF). Firstly, a signed pressure force function is given based on image statistic information, which can segment the gray uneven images, and a weighted function of the global gray mean of the inside and the outside of contour curve is developed by combining the internal with the external gray mean values. Another signed pressure force function is defined by using image global information to segment the weak boundary images. Then, by combining the above two proposed signed pressure force functions, i.e. the signed pressure force function of statistic information and the other function of the global information(called as double signed pressure force functions), a new evolution equation of level set is designed by adding the combined weighting coefficients. Finally, DSPF is introduced in the selective binary and Gaussian filtering regularized level set model, and a novel active contour image segmentation algorithm based on DSPF (DSPF-ACIS) is proposed. Experimental results show the advantages of the proposed method in segmenting the weak boundary and gray uneven images in terms of both efficiency and anti-noise performance.

Key words: active contour, signed pressure force function, level set;image segmentation

摘要: 为了解决测地线模型和CV模型无法同时对弱边界、灰度不均匀图像进行分割的问题,提出一种基于双符号压力函数的活动轮廓图像分割方法。首先,基于图像统计信息定义分割灰度不均匀图像的符号压力函数,基于内部和外部灰度均值给出轮廓曲线内外的全局区域灰度均值的加权组合函数,运用图像全局信息定义分割弱边界图像的符号压力函数;然后,结合统计信息的符号压力函数和全局信息的符号压力函数(简称“双符号压力函数”),通过增加组合的权值系数,设计新的水平集演化方程;最后,将双符号压力函数引入到二值选择和高斯滤波正则化水平集模型中,构建一种基于双符号压力函数的活动轮廓图像分割算法。仿真实验结果表明,该算法能够有效地分割弱边界、灰度不均匀的图像,同时对噪声也有一定的抗干扰性。

关键词: 活动轮廓, 符号压力函数, 水平集, 图像分割