An Efficient Hierarchical Segmentation Approach for Remote Sensing Images

摘要

Segmentation is one of the key problems in remote sensing image processing, and a cluster hierarchy is a powerful tool to analyze data on multiple scales. In this paper, we present an optimal hierarchical remote sensing image segmentation approach which is a combination of feature-based classification, multi-scale region merging and classification-based object refinement, and it is shown to have the following properties: 1) it gets better results on general remote sensing images; 2) it offers a flexible segmentation hierarchy at a less computational cost.