Edge-guided segmentation method for multiscale and high resolution remote sensing image

In order to overcome the complexity of region merging in the segmentation of high resolution remote sensing images, an edge-guided segmentation method for multi-scale and high resolution remote sensing image was proposed. First, SUSAN operator was used to extract feature edges from the original test image. Then, a graph-based segmentation algorithm was used in the first-stage image segmentation and the following region merging stage, and the extracted edges were efficiently used to guide merging process. To validate the proposed method, two experiments were performed on QuickBird image. The results were compared with the segmentation results of eCognition and method without edge-guide. The results show that this proposed method can efficiently depress the region merging in low-contrast areas for the traditional image segmentation algorithms, and make it possible to choose a reasonable segmentation scale in the whole image.