Computer Engineering and Applications ›› 2020, Vol. 56 ›› Issue (7): 228-233.DOI: 10.3778/j.issn.1002-8331.1910-0323

Previous Articles     Next Articles

Active Contour Image Segmentation Combined with Statistical Modeling of Distribution Metrics

LU Yuanyuan, FENG Hao, LI Jing   

  1. 1.School of Information Engineering, Wuhan College, Wuhan 430212, China
    2.National Engineering Research Center for E-Learning, Central China Normal University, Wuhan 430079, China
  • Online:2020-04-01 Published:2020-03-28

结合分布度量统计建模的主动轮廓图像分割

鲁圆圆,冯浩,李靖   

  1. 1.武汉学院 信息工程学院,武汉 430212
    2.华中师范大学 国家数字化学习工程技术研究中心,武汉 430079

Abstract:

Image segmentation is an indispensable key step in digital image processing. In order to solve the problem that the traditional active contour model is inaccurate for the non-homogeneous image segmentation and the segmentation efficiency is low, this paper proposes an active contour image segmentation algorithm combined with statistical modeling of distribution metrics. The energy driving force of the proposed algorithm takes into account the global statistical modeling information of the image and other mixed gray distribution information, so that the segmentation curve can be more accurately evolved to the target edge. Firstly, the distribution metric energy driving force is defined as the variance of the ratio distance defined by the probability density function inside and outside the contour. The energy driving force is based on the statistical modeling of the image global information, which can more accurately describe the energy variation inside and outside the contour curve. Secondly, the mixed gray-scale distribution energy driving force is represented by the L2 norm of the image fitting value center of the image gray value and the fusion mean and median. Finally, the distribution metric energy driving force and the mixed gradation distribution energy driving force are combined to form a novel energy function, and the minimum value of the energy function is obtained by using the level set method and the gradient descent method to obtain the final image segmentation result. Compared with the image segmentation experiment results of four algorithms, such as traditional CV(Chan Vese) model and LBF(Local Binary Fitting) model, the proposed model has great advantages in the subjective evaluation effect, sensitivity to the initial contour, running time and number of iterations.

Key words: image segmentation, active contour model, global statistical modeling information, mixed gray distribution information, energy function

摘要:

图像分割是数字图像处理中不可或缺的关键步骤。为了解决传统主动轮廓模型针对非匀质图像分割结果不准确且分割效率低的问题,提出一种结合分布度量统计建模的主动轮廓图像分割算法。所提算法的能量驱动力兼顾了图像的全局统计建模信息和其他混合灰度分布信息,使得分割曲线能够更加精确地演化至目标边缘。分布度量能量驱动力定义为轮廓内外概率密度函数定义的比率距离的方差,该能量驱动力基于图像全局信息统计建模,能够更加精确地描述轮廓曲线内外的能量变化;混合灰度分布能量驱动力由图像灰度值与融合均值与中值的区域拟合中心的L2范数表示。将分布度量能量驱动力与混合灰度分布能量驱动力组合形成新的能量泛函,利用水平集方法和梯度下降法迭代求得该能量泛函的最小值,以获得最终的图像分割结果。与传统CV(Chan Vese)模型、LBF(Local Binary Fitting)模型等四种算法的图像分割结果相比,所提模型在主观视觉效果、对初始轮廓的敏感性、运行时间和迭次次数方面均具有较大优势。

关键词: 图像分割, 主动轮廓模型, 全局统计建模信息, 混合灰度分布, 能量泛函