提出基于图像内容层次表征的高分辨率遥感图像快速多精度分割方法。首先根据初始分割结果建立区域邻接图(RAG),并将其定义为马尔可夫随机场(MRF);然后引入光谱、形状和边缘等图像特征进行层次合并,通过记录层次合并过程获得图像内容的层次表征;最后根据层次表征中不同层级对象之间的关系快速生成任意不同精度的分割结果,以满足不同应用的需求。利用QuickBird卫星图像进行实验和评价的结果表明,本文方法具有较高的精度和效率。

A segmentation method for high-spatial resolution remotely sensed images segmentation based on the hierarchical representation of the image content is proposed in this paper.The hierarchical representation is built through a hierarchical merging process based on an initial segmentation.First,we build the region adjacency graph (RAG) on which a Markov random field (MRF) is defined.Then the hierarchical merging process,where the merging criterion includes multi-spectral,edge and shape features,is applied and recorded on the RAG to obtain the hierarchical representation.It is quite efficient to produce a number of segmentation results with different precisions using the relationship of different objects in the hierarchical representation.Experimental results on QuickBird images show that the proposed method can greatly improve the segmentation efficiency and produce high quality results.