摘要
提出了一种改进的变分水平集分割算法。引入演化曲线内外灰度图像的中值代替传统CV(Chan_Vese)模型中的均值作为曲线拟合中心。在轮廓初始化后,采用最大类间方差(OTSU)方法改进曲线拟合中心,将最大类间方差与拟合中心结合,提高分割的准确率和适应性。加入双阱势的距离正则项以避免水平集重新初始化,提高效率,从而得到一个自适应阈值与区域信息相结合的水平集演化模型。
A segmentation algorithm based on an improved variational level set is proposed.The mean value of gray image inside and outside the evolution curve is introduced to replace the mean value in the traditional CV(Chan Vese)model as the curve fitting center.After the contour initialization,OTSU method is used to improve the curve fitting center,which combines the maximum inter class variance with the fitting center to improve the accuracy and adaptability of segmentation.The distance regularization term of double well potential is added to avoid the reinitialization of level set and improve the efficiency so that a level set evolution model combining adaptive threshold and regional information is obtained.
作者
温佳
杨杰伟
杨亚楠
WEN Jia;YANG Jiewei;YANG Ya’nan(College of Electrical and Information Engineering,Tianjin Polytechnic University,Tianjin 300387,China)
出处
《实验技术与管理》
CAS
北大核心
2020年第1期71-74,共4页
Experimental Technology and Management
基金
国家自然科学基金项目(61401439,61601323)
天津市自然科学基金项目(17JCQNJC01400).
关键词
图像分割
水平集演化
类间方差
双阱势距离正则项
image segmentation
level set evolution
inter-class variance
regular term of double well potential distance