基于图割和分水岭变换的图像分割方法  

Image segmentation method based on graph cut and watershed transformation

作　　者：李海滨[1] 杜益福[2] 刘彬[1] 

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004 [2]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《燕山大学学报》2013年第2期143-147,共5页Journal of Yanshan University

基　　金：国家自然科学基金资助项目(61007003);河北省科学技术研究与发展计划项目(10213527);高等学校博士学科点专项科研基金(20101333120006)

摘　　要：为了解决分水岭变换在图像分割中的过分割现象和提高图割在图像分割上的速度,提出了一种基于图割和分水岭变换相结合的图像分割算法。首先对图像进行分水岭变换,将图像分割成多个小区域;然后利用每个区域内的像素平均信息代替每个像素信息,构造简化的网络图;最后利用图割算法实现网络图分割,产生最终分割结果。实验结果表明:本文算法既能解决分水岭变换所产生的过分割现象,又能提高图像分割的速度。An image segmentation algorithm based on the graph cut and watershed transformation is proposed in this paper to solve the watershed transformation over-segmentation phenomena and improve the speed of graph cut.First,the watershed transformation is applied to segment the image into a large number of small partitions.Then the simplified network graph is constructed using the average information of the pixels.Finally,the graph cut algorithm is employed to realize network graph segmentation and get the segmentation results.The experiment results indicate that the proposed algorithm can not only solve the watershed transformation over-segmentation phenomena,but also improve the speed of the image segmentation algorithm.

关 键 词：图割 分水岭变换 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]