基于改进的分水岭算法图像分割方法研究  

Image Segmentation Based on Watershed Algorithm

作　　者：徐奕奕[1] 刘智琦[1] 刘琦[1] 

机构地区：[1]广西工学院计算机工程系,广西柳州545006

出　　处：《计算机仿真》2011年第9期272-274,共3页Computer Simulation

基　　金：广西科学研究与技术开发计划项目(桂科研0992006-13)

摘　　要：研究图像分割中保真问题,针对目前分水岭算法易产生过分割现象,导致图像分割边缘不明显现象,使得分割以后图像失真较大。为了解决上述问题,提出了一种基于空间模式聚类算法和分水岭算法相结合的图像分割方法。首先对图像采用分水岭算法进行图像分割,然后对产生过分割现象的图像区域采用基于空间模式聚类方案进行合并。仿真结果表明,与传统的分水岭图像分割方法相比,分割出的图像边缘效果明显清晰,有效保证原图效果。In order to solve the problem of over-segmentation existing in traditional watershed algorithm, a new image segmentation algorithm based on watershed algorithm and clustering procedure of spatial patterns was proposed. It first used watershed algorithm for the image segmentation, defined the characteristic quantities for each watershed region, and then merged the small image regions caused by the over-segmentation using the method based on the spa- tial pattern clustering. Compared with the traditional watershed method, the image edges are more clear. Image seg- mentation experiment results proved the feasibility and effectiveness of the algorithm.

关 键 词：图像分割 空间模式聚类 分水岭 隶属度 欧式距离 空间距离 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]