A REMOTE SENSING IMAGE SEGMENTATION FOR EXTRACTING HIGH SLOPES

An object-oriented image segmentation algorithm combining edge detection and region segmentation is proposed to segment color petro-graphic images. First, Susan Operator is applied to obtain the edge map, and then a two-stage region segmentation method is used to get the partition. In the first phase, quad tree structure groups pixels to form image object/region primitives. At the beginning of the second phase, the first phase primitive partition is converted into regions with concrete features, such as boundary length, neighbor relations and then a hierarchical region merging operation is performed with those primitives, where the criterion for region merging integrates edge map and other region features. This segmentation algorithm has been tested on the QuickBird image of Yunyang county, and experiment results demonstrate stable convergence and good efficiency.