基于自适应局部阈值的彩色图像分割  

Color Image Segmentation Based on Adaptive Local Thresholds

作　　者：冉玉梅[1] 王洪国[1] 杨玉会[1] 许镇[2] 

机构地区：[1]山东师范大学管理与经济学院,山东济南250014 [2]山东师范大学信息科学与工程学院,山东济南250014

出　　处：《计算机与现代化》2010年第1期109-112,共4页Computer and Modernization

基　　金：山东省自然科学基金资助项目(Q2006G03)

摘　　要：使用基于图论的最小生成树方法进行图像分割,能够获得全局的分割效果及良好的运行效率,但是图中边的数目太多,而且使用固定阈值合并的最小生成树不能适用于所有彩色图像。本文针对这一问题,结合分水岭方法,先将分水岭变换产生的初始分割区域转变成图论中的顶点和边,再用最小生成树合并,通过分析局部信息变化,得到自适应局部阈值。实验表明,基于自适应局部阈值的最小生成树分割效果比基于全局阈值最小生成树的分割效果要好。The using of minimum spanning tree based on graph theory can obtain global division effect and good operating efficiency. But the number of arcs is too large, and the method through fixed global threshold doesn＇ t adapt to all color images. This paper aims at these problems, combines the minimum spanning tree method with watershed algorithm, firstly, transform the initial segmentations into nodes and arcs of graph, then merge them through minimum spanning tree. By analyzing the local changes of the regions characteristics, adaptive local thresholds are derived. The experiment indicates that, the minimum spanning tree method based on adaptive local thresholds is better than the minimum spanning tree based on global threshold.

关 键 词：彩色图像 分水岭 最小生成树 自适应局部阈值 图论 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]