摘要
文章对最大类间方差法进行了改进,利用计算图像类间方差的约束条件,通过构造罚函数,把求最大类间方差问题转换为求带约束条件的函数最优解问题,再用迭代法对图像进行分割,从而有效地弥补了最大类间方差法对图像噪音和目标大小敏感不足。仿真实验结果表明,该方法效果良好。
Aim.The shortcoming of the widely used Otsu′s method of maximum between-cluster variance[5] is that,in our opinion,it is highly sensitive to image noise and size of target.We propose improving it through suppressing such sensitivity.Section 1 of the full paper briefs N.Otsu′s method.Section 2 explains in some detail our improvement on Otsu′s method;its core consists of：（1） using the constraints on calculating image between-cluster variance and using the penalty function proposed by us,we convert the problem of counting maximum between-cluster variance into that of obtaining the optimal value of function with constraints;（2） the iterative method that performs image segmentation suppresses much the sensitivity to image noise and target size.Our improvement on Otsu′s method is proved effective preliminarily by experimental results as presented in Figs.1 and 2.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2010年第2期259-263,共5页
Journal of Northwestern Polytechnical University
基金
航空科学基金(20090153002)资助
关键词
目标提取
图像分割
类间方差
extraction
image segmentation
target extraction
between-cluster variance
