A multi-scale region growing segmentation for high resolution remotely sensed images

For remotely sensed images, region growing segmentation is a widely-used technique for object extraction and identification. However, the main drawback of region growing segmentation is the threshold setting for stopping the growth of a region. Improper threshold setting not only causes over-segmentation or under-segmentation, but also cannot create needed object regions for various targets. In order to overcome these drawbacks, a multi-scale region growing segmentation method, based on the maximization of the change of edge density, is proposed. Experiments, including different kinds of high resolution remotely sensed images are used to test the performance of the proposed scheme. The experimental results show that the proposed scheme can both remove the limitation of threshold setting and generate relatively reasonable segmentation output for different types of objects.