华东师范大学学报(自然科学版) ›› 2014, Vol. 2014 ›› Issue (6): 57-66.doi: 10.3969/j.issn.1000-5641.2014.06.009

• 计算机科学与技术 • 上一篇    下一篇

结合区域间差异性的水平集演化模型

陈雯,朱敏   

  1. 华东师范大学 信息科学技术学院计算中心, 上海 200062
  • 出版日期:2014-11-25 发布日期:2015-02-07
  • 通讯作者: 朱敏, 女, 高级工程师, 硕士生导师,研究方向为图像处理与模式识别 E-mail:mzhu@cc.ecnu.edu.cn
  • 作者简介:陈雯, 女, 硕士研究生, 研究方向为图像处理与模式识别.E-mail: chenwen251314@yeah.net

Level set evolution model with inter-region dissimilarity

 CHEN  Wen, ZHU  Min   

  1. The Computer Center, School of Information Science Technology, East China Normal University, Shanghai 200062, China
  • Online:2014-11-25 Published:2015-02-07

摘要: 水平集方法在图像分割中得到了广泛的应用.其中基于边缘的活动轮廓模型主要通过梯度信息驱动曲线演化到目标边界,但基于梯度信息使得在分割时会产生过分割,
并且对于灰度不均匀图像处理效果不理想, 有可能得到不令人满意的结果. 而基于区域的活动轮廓模型则是通过区域信息控制曲线移动,使得分割的结果立足于整体图像信息. 基于上述原因,本文通过在水平集中提出了一种新的区域分量,在能量泛函中加入目标区域灰度和背景区域灰度的差的平方,提出了一种改进的图像分割算法. 研究结果表明,与一般的活动轮廓模型相比,加入区域间差异性信息的活动轮廓模型的分割结果更加符合实际情况并且收敛速度更快,效率更高, 得到的分割结果更令人满意.

关键词: 区域间差异性, 水平集方法, 自适应变化系数, 几何活动轮廓模型

Abstract: Level set method has been widely used in image segmentation. Edge-based active contour model mainly drives the curve to the target boundary by gradient information, but the model based on gradient information makes the segmentation produce over-segmentation. And for the uneven gray image, the processing results are not so satisfactory.\linebreak However, the region-based active contour model controls the curves evolution through regional
information, which will get the segmentation results based on the whole image. For the reasons stated above new regional component is put forward in the Level set, by utilizing the square of mean gray value difference between the background and the target area in energy function, a new improved method for image segmentation is proposed. Compared with the general active contour model, the experimental results show that the active contour model~with the
regional difference information has ideal effect of segmentation,~faster evolution speed and higher efficiency. We can get more satisfactory segmentation results.

Key words: inter-region dissimilarity, level set method, adaptive evolution coefficient, geometric active contour model

中图分类号: