摘要
Mean Shift算法分割图像时,带宽的大小直接影响分割效果。带宽分为空间带宽和值域带宽。本文根据待分割遥感图像的空间分辨率参考选定空间带宽,基于渐近积分均方差最小原则计算每一波段值域带宽;针对MS算法分割图像时存在过分割问题,提出基于区域面积加权的区域相似度准则和基于区域熵的合并停止准则来合并分割后区域。MATLAB软件3组实验结果表明:本文方法相比EDISON软件能得到更好的分割效果,且能在一定程度上提高遥感影像分割的自动化。
Bandwidth is a key parameter in Mean Shift based image segmentation. Bandwidths include spatial bandwidths and range bandwidths. Spatial bandwidths are picked according to remote sensing image' s spatial resolution. Gray range bandwidths are calculated by Minimizing Asymptotic Mean Integrated Squared Error ( AMISE). Segmented regions is merged by using regions areas weighed similarity rule and region entropy based region merge cease rules to solve over-segmentation of MS algorithm. MATLAB software experiment results of three images showed that this method has better segmentation results than EDISON software, and in some extent could improve remote sensing image auto-segmentation level.
出处
《测绘科学》
CSCD
北大核心
2012年第6期98-100,106,共4页
Science of Surveying and Mapping
基金
国家高技术研究发展计划(863计划)重点项目(2009AA122004)
国家自然科学基金青年科学基金项目(40901171)