高斯光谱估计结合形状特征的遥感影像分割方法

Segmentation Method of Remote Sensing Images Based on Gaussian Estimates and Shape Characteristics

针对传统区域合并算法仅利用光谱均值的不足,提出了一种新的结合光谱与形状特性的遥感影像分割算法。该算法首先对输入遥感影像进行分水岭分割;然后利用一种新的结合光谱与形状特性的区域相似性准则,采用区域邻接图的算法对初始分割区域进行合并,获取最终分割结果。实验结果表明,该算法能有效克服传统算法仅利用光谱均值的不足,有效解决遥感影像中相同光谱,不同纹理结构区域的分割问题;同时,本文算法还对纹理规则区域的分割效果有较好的改进。

With the insufficiency of only using the average spectrum in traditional region merging algorithm, a new remote sensing im- age segmentation algorithm which combines the characteristics of spectral and shape is proposed. Firstly, the algorithm segments the entering remote sensing image, and then uses a new regional similarity criteria which combines the characteristics of spectral and shape and region adjacency graph algorithm to merge the initial segmentation area. Lastly the final segmentation results are obtained. Experi- mental results show that the algorithm can effectively overcome the insufficiency of only using the average spectrum in the traditional al- gorithm and effectively solve the segmentation problem of remote sensing image in the same spectral but different texture regions. At the same time, the segmentation results of the rule texture area have been improved.