基于Mean Shift聚类的多级阈值化方法  

Based on the Mean Shift Clustering Multilevel Threshold Method

作　　者：程全 刘晓青 刘玉春 王志良[2] 

机构地区：[1]周口师范学院机械与电气工程学院,河南周口466001 [2]北京科技大学计算机与通信工程学院,北京100083

出　　处：《郑州大学学报（工学版）》2017年第6期64-69,共6页Journal of Zhengzhou University（Engineering Science）

基　　金：国家自然科学基金资助项目(61401526);河南省自然科学基金资助项目(152300410134)

摘　　要：为了解决多级阈值化技术中所选阈值的数量通常不能预先确定的问题,提出一种基于Mean Shift聚类技术的新型多级阈值化方法.首先,通过使用Mean Shift技术探寻出潜在的模式中心,应用迭代的阈值选择方法来自动确定相邻模式中心的各个阈值;然后,采用多级阈值化对图像进行分割;最后,通过实验验证了基于Mean Shift聚类技术分割的图像相对于原始图像的对比度有了很大提高.该方法通过简单修改程序参数就能够灵活控制分割精度,可以广泛应用于单阈值分割、多级阈值分割和有损压缩等技术中.In order to solve the problem that the number of selected thresholds in multilevel thresholds cannot be usually predetermined,a novel multi-level thresholding method based on Mean Shift Clustering technique was proposed.Using Mean Shift technology to explore the potential mode center,the various thresholds of ad-jacent to the mode center was automatically determined by using of iterative threshold selection method, and then the method of multi-level threshold was used for image segmentation.The experimental results showed that,relative to the original image,contrast of the image split with Mean Shift clustering technique was greatly improved.This method could control the segmentation precision flexibly by simply modifying parameters of the program,and could be widely used in the technology of single threshold segmentation, multi-level threshold segmentation and detrimental compression and other technologies.

关 键 词：多级阈值化 图像分割 迭代阈值化 分割质量评估 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]