基于ROI信息的活动轮廓模型MR图像分割  

MR Image Segmentation by Active Contour Based on Information of the Region of Interesting

作　　者：张建伟[1] 陈允杰[1] 朱玉辉[1] 夏德深[2] 

机构地区：[1]南京信息工程大学数学系,南京210044 [2]南京理工大学计算机系,南京210094

出　　处：《计算机工程与应用》2006年第1期37-39,63,共4页Computer Engineering and Applications

基　　金：香港特区政府研究资助局研究项目(编号:CUHK/4180/01E;CUHK/1/00C)

摘　　要：参数活动轮廓模型和几何轮廓模型是目前常用的图像分割方法。传统的参数活动轮廓模型在曲线演化过程中仅使用了梯度信息,在处理具有强噪音或弱边界的目标时往往得不到理想的结果,且不具有拓扑可变性。为此,构造一种基于区域信息约束力的参数活动轮廓模型,在新的约束力的作用下,模型可以有效降低噪音的影响并防止曲线从弱边界泄漏,且提高曲线进入细窄区域的能力,将参数模型几何化,使该模型具有拓扑可变的能力。实验表明,该模型具有较好的分割效果。In recent years,the fields of active-contour based image segmentation have seen the emergence of two competing approaches,The first is based on parameter models,the other is based on geography models.Parameter models used to use information of gradient,so it is sensitive to the strong noise and pass through the weak edge easily.It can not deal with topologic change during the segmenting.In this paper information of the region of interesting is used to make pressure force to reduce the effect of the noise and weak edge.As to make the model have the ability to change topological structure,the active model is converted to geography model.Experiments show the new model has better results.

关 键 词：SNAKE模型 LEVEL Set模型 区域信息 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]