Experimental Study on Graph-Based Image Segmentation for the Feature Extraction of High-Resolution Satellite Image

摘要

The development of image segmentation algorithms for remote sensing images has been greatly increased after the availability of high-resolution imagery expanded. Graph-based segmentation techniques have many advantages because they utilize both boundary and regional information without any pre-known information. Despite these advantages, graph-based methods are generally not used for remote sensing, because of the long computational time and complexity of the calculations. In this paper, we aim to evaluate MST graph-based image segmentation for high-resolution satellite images and improve their performance with shape features.