基于改进水平集和区域生长的轮廓提取方法  

Contour extraction method based on improved level set and self-adaptive region growing

作　　者：付英杰[1] 张剑[1] 邹翎[2] 王凯[1] 李柏林[1] 吕曦[2] 宋思思[2] 

机构地区：[1]西南交通大学机械工程学院,成都610031 [2]四川大学华西医院放射科,成都610041

出　　处：《计算机应用研究》2012年第7期2770-2772,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(30800263);四川省科技支撑科研资助项目(2010GZ0187);中央高校基本科研业务费专项资金资助项目(2010XS16)

摘　　要：以人体大脑海马切片序列为研究对象,提出了基于目标灰度差异的水平集方法,该方法克服了水平集方法用于提取海马轮廓时边界易停留在背景梯度局部极值处的问题。同时为了避免传统区域生长人工定义阈值的盲目性,将水平集分割后的图像的标准差作为阈值进行自适应区域生长,获得了海马的轮廓。通过理论分析与实验验证,提出的方法能够有效地滤除非目标区域对海马目标的干扰,获得了较好的分割效果。This paper proposed a new level set based on target gray differences to research the human brain hippocampal slice sequences.It solved the boundary staying at the background’s extreme gradient in traditional methods.In order to avoid the blindness in region growing,it put forward the processed image’s standard deviation as the threshold.Then it received the hippocampus’s contour.By theoretical analysis and experiment verification,this method can remove non-target parts from target efficiently,and a better segmentation result is obtained.

关 键 词：海马 水平集方法 自适应区域生长 阈值 轮廓提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]