基于分水岭和α扩展的遥感影像分割方法研究

Research of Remote Sensing Image Segmentation Based on Watershed and Alpha Expansion

提出了一种新的影像分割方法。方法利用Meanshift算法对原始影像进行预处理,再运用分水岭标记算法获得初始分割结果,得到独立的影像斑块。以每一个斑块作为一个结点构建一幅图,应用图割理论中的α扩展原理对所构建的图进行多类分割。由于充分利用分水岭算法和α扩展理论的优点,新方法分割结果具有更高的准确性,并有相应的实验证明该方法的有效性。

A novel image segmentation method is proposed.It firstly uses meanshift algorithm to go on pre-processing to the original image,then use marked watershed to get the initial segmentation results which can obtain independent image patches.Each patch as a node to constructs a graph,then applicate the alpha expansion principle in the graph cut theory to the constructed graph to go on the multi class segmentation.Because of the full use advantages of watershed algorithm and alpha expansion theory,the segmentation results of the new method have higher accuracy,and the corresponding experiments have proved the new method＇s effectiveness.