计算机工程

• 图形图像处理 • 上一篇    下一篇

水平集图像分割核映射模型的改进

张迎春1,郭禾1,张丹枫2   

  1. (1.大连理工大学软件学院,辽宁 大连116620; 2.大连大学医学部,辽宁 大连116622)
  • 收稿日期:2015-05-29 出版日期:2016-04-15 发布日期:2016-04-15
  • 作者简介:张迎春(1980-),女,博士研究生,主研方向为图像分割、粗糙集理论;郭禾,教授;张丹枫,硕士研究生。
  • 基金项目:
    国家自然科学基金资助项目(61033012)。

Improvement of Kernel Mapping Model for Level Set Image Segmentation

ZHANG Yingchun 1,GUO He 1,ZHANG Danfeng 2   

  1. (1.School of Software Technology,Dalian University of Technology,Dalian, Liaoning 116620,China;2.Department of Medicine,Dalian University,Dalian,Liaoning 116622,China)
  • Received:2015-05-29 Online:2016-04-15 Published:2016-04-15

摘要: 为提高水平集演化迭代效率和图像分割质量,对核映射(KM)模型进行改进,依据粗糙集理论提出图像数据离散化的定义和方法。由于加权后的KM模型推导出的区域参数具有收敛性,且离散化后图像灰度均匀性有所提高,区域参数可以更好地表达对应演化区域的灰度值,使得图像能够被更精确地分割。在迭代演化过程中,加权后的KM模型对应的水平集权重最小值为1。权重值越大水平集元素更新与演化曲线收敛越快,原始KM模型对应的非加权水平集可看作是所有权重值为1的加权水平集,提出的方法能够以相对较高的迭代效率完成图像分割。合成图像和自然图像分割实验结果表明,改进后的水平集图像分割方法有更好的分割质量和迭代效率。

关键词: 图像分割, 水平集, 离散区域, 灰度不均匀, 图像数据离散化, 粗糙集

Abstract: In order to increase iterative efficiency and image segmentation quality,an improved Kernel Mapping(KM) model is presented.Both definition and method of image data discretization are proposed based on the rough set theory.The image after discretization will enhance intensity homogeneity and every region parameter sequence deduced by the weighted KM model converges,so that the region parameters can better express the gray value among the evolving regions.Therefore,the image can be more accurately segmented.During each iteration,the minimum of the weighted value of the level set elements is one according to the weighted KM model.The larger value for the weight becomes,the faster the level set element is updated and the evolving curve converges.The level set of KM model without weighted can be taken as the level set which the weighted value of every element equals one.Thus the weighted KM model can fulfill an image segmentation with high efficiency.Experimental results on synthetic and natural images show that the proposed method has better segmentation quality and iteration efficiency.

Key words: image segmentation, level set, discrete region, intensity inhomogeneity, image data discretization, rough set

中图分类号: