Abstract: A novel method for range image segmentation is presented in this paper, It is based on an integration of edge and region information. The algorithm consists of three steps: edge and critical point detection, triangulation, and region information. The edge detection method presented in this paper is based on morphological operations. In general, segmentation may not be effective when only edge operators are applied on range images especially on noisy images. Further processing is important for final segmentations when the edge operators are not sufficient. In this paper, critical points are extracted from planar edge curves. These edge curves and critical points constitute an initial set of segments. The constrained Delaunay triangulation is employed on the initial set to obtain triangle-like connection graphs. By projecting the critical points and their connectivity relationships in parallel onto 3D surface, a 3D surface structure graph (SSG) is obtained. Hence, segmentation is completed by grouping these triangle-like facets. The grouping scheme is presented in this paper according to the normals of adjacent facets. Because edge curves are not usually straight lines but rather a set of curve segments, we introduce extensive triangulation for building 3D triangle-like surface structure graphs (SSG's). This method significantly reduces the computation complexity compared to polyhedral approximations using the original Delaunay triangulation. Experimental results show that the method is efficient for range image segmentation especially for polyhedra.!1
展开▼
