摘要
人脑MR图像中的海马结构存在低对比度、边界模糊等缺点,给海马的轮廓分割带来较大干扰。为解决水平集分割海马时边界容易停留在非目标区域梯度极值处的问题,提出一种改进的水平集方法。从图像全局出发考虑方差信息,在水平集函数的外部能量泛函中增加波动能量项,驱动零水平集曲线向灰度波动较小的区域运动。实验结果表明,该方法可提取出MR图像中的海马轮廓,分割效果较好,演化速度有所提高。
The contour extraction of hippocampus in human brain MR images has a lot of interference which is caused by its low contrast, fuzzy boundaries, and so on. To solve this problem that the boundary stays at gradient extremal in non-target area when level set segmentation method is applied, in this paper, a method based on improved level set is proposed. It considers the variance information through the whole image, adds wave energy items in the external energy functional of tile level set function, and drives the zero level set curve to where gray-scale fluctuation is smaller relative to the hippocampus. Experimental results show that this method can ideally extract the contour of the hippocampus from the MR images with excellent segmentation results and greatly improves the evolution speed.
出处
《计算机工程》
CAS
CSCD
2013年第6期283-286,共4页
Computer Engineering
基金
四川省科技创新苗子工程基金资助项目(2011-001)
四川省成都市科技计划基金资助项目(11PPYB109SF)
关键词
人脑海马
水平集方法
梯度极值
灰度波动
全局方差
轮廓提取
human hippocampus
level set method
gradient extremum
gray-scale fluctuation
global variance
contour extraction
