文档介绍：第 24 卷第 2 期北京理工大学学报 VoI. 24 No. 2
2004 年 2 月 Transactions of Beijing Institute of TechnoIogy Feb. 2004
文章编号:1001-0645(2004)02-0162-04
对复杂边缘检测的 Snake 改进算法
赵保军, 李栋
(北京理工大学信息科学技术学院电子工程系,北京 100081)
摘要:针对 Snake 原始模型对包含凹陷形状目标或信噪比较低的图像边缘检测结果差的问题,通过对复杂形状
目标及含噪声图像的特性研究,从能量函数和迭代策略两个方面对 Snake 原始模型进行改进,增加了新的控制能
量,并对各能量因素的影响权值进行了讨论,同时对蛇点采用了动态分布方法,以适应不同的目标形状特性。仿真
结果表明,改进后的 Snake 模型较原始模型大大减弱了对蛇点初始位置的依赖,并在一定程度上有效地克服了图像
噪声对迭代算法的影响,提高了对复杂目标的边缘检测性能。
关键词:Snake 模型;图像分割;能量函数;边缘检测
中图分类号:TN 911. 73 文献标识码:A
Effective Boundary Detection Method plex
Target Based on the Snake Model
ZHAO Bao-jun, LI Dong
(Department of EIectronic Engineering,SchooI of Information Science and TechnoIogy,Beijing Institute of
TechnoIogy,Beijing 100081,China)
Abstract:When image objects have non-convex forms or when there exists Iower signaI-noise rate in the
image,the traditionaI Snake modeI often cannot detect the object edge effectiveIy. The articIe attempts to
study the characters of image that pIex shape target or noise,so as to improve the traditionaI
Snake in the energy function and operation strategy. It gives new controI energy function and discusses
the right vaIues of these functions,meanwhiIe dynamicaIIy arranges the Snake points for different target
forms. SimuIation proves that the new Snake modeI decreases greatIy the Snake*s dependence on initiaI
conditions and es