Edge-Guided Segmentation Method for Multi-Scale High Resolution Remote Sensing Image
一种边界引导的多尺度高分辨率遥感图像分割方法

In order to overcome the complexity of region merging in image segmentation, a method of edge-guided multi-scale high resolution remote sensing image segmentation is proposed. SUSAN operator is first used to extract feature edges from the original test image, then an efficient graph-based segmentation algorithm is realized in the first-stage image segmentation and in the following region merging stage, extracted edges are efficiently used to guide merging process. To validate the proposed approach, two experiments were performed on QuickBird image, which compared segmentation results among edge-guiding method, without edge-guiding method and eCognition, and results show that this proposed method can efficiently solve the problem of region merging in low-comparison areas, and the limitation of not being able to choose a reasonable segmentation scale which apply to the whole image is also eliminated in this proposed method.