改进的分水岭分割算法在遥感图像分割中的应用  

Application of remote sensing image segmentation based on improved watershed segmentation algorithm

作　　者：顿德光[1] 张绍良[1] 

机构地区：[1]中国矿业大学环境与测绘学院,江苏徐州221008

出　　处：《企业技术开发》2010年第11期4-5,共2页Technological Development of Enterprise

基　　金：国家科技支撑项目"应急实时灾区数据快速获取与处理技术研究"(2008BAK49B02);中国测绘科学研究院重大科技专项"测绘遥感数据快速获取与处理集成系统"(7771016)

摘　　要：为了解决分水岭算法造成的过度分割及二值化阈值选择问题,文章提出了一种新的算法。首先利用数学形态学算法对图像进行增强处理,接着用Sobel算子计算获得梯度图像并设置一阈值,然后合并小于一定面积的区域后进行分水岭分割,最后根据阈值选择算法选取分割阈值,在此基础上二值化得到感兴趣目标。实验结果表明,该方法能很好的抑制过度分割,快速的获取感兴趣目标。This article raises a new algorithm which is proposed to solve the problem of threshold selection and the over-segmentation caused by watershed segmentation algorithm.First,the image is enhanced based on mathematic morphology.Then the gradient image,which is calculated by the Sobel operator is used for over-segmentation and a threshold is set.And then the certain area is chosen for different kind of image,so it leads to the homogeneous regions in the segmentation result.Finally the interest target is received based on threshold selection algorithm.The experimental results show that this method can be powerful to restrain the over-segmentation and the interesting target can be fast received.

关 键 词：数学形态学 分水岭分割算法 遥感图像 均值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]