基于最近邻有向图的遥感图像快速分割算法

Fast Remote Sensing Image Segmentation Algorithm Based on Nearest Neighbor Direct Graph

针对现有的区域生长算法没有考虑到区域之间最近邻关系的有向性这一问题,提出了基于最近邻有向图的遥感图像快速分割算法。首先使用分水岭算法对遥感图像进行初次分割,然后在分割得到的区域对象基础上建立最近邻有向图。在区域生长过程中,沿着有向边形成的路径合并相邻的区域对象。当所有合并完成后重构区域对象的最近邻有向图,进行下一轮合并,直至区域数目不再变化。该方法避免了每次合并一个区域对象就重新计算新的邻居关系,从而降低了计算复杂度。实验结果表明,该方法分割结果比较合理,与其他几种方法相比运行效率明显提高。

The existing region growing algorithms do not take into account the direction of the nearest neighbor rela-tions, which results in frequent rebuilt of the neighbor relations. In this paper,a fast algorithm for remote sensing image segmentation was proposed based on nearest neighbor directed graph. First of all,a remote sensing image was segmen- ted using the watershed algorithm, and then a nearest neighbor directed graph was established on the basis of the region objects of the previous segmentation. In the region growing phrase, the adjacent region objects were merged along the directed edges. When the first round is finished, the nearest neighbor directed graph should be rebuilt, and the second round of region growing is initiated. This process repeats until the region number is no longer changed. This method a- voids recalculating the neighbor relations whenever a merge happens, which reduces the computational complexity. The experimental results show that the algorithm proposed in this paper is more reasonable, more efficient compared with the other three algorithms.