Hierarchical segmentation of range images inside the combinatorial pyramid

摘要: RGB-D cameras are not only able to provide color (Red-Green-Blue -RGB-) information from the scene but also a relatively accurate cloud of 3D points. Using coming this organized cloud, it is possible define around each image pixel small planar patch and obtain its normal vector. Within framework combinatorial pyramid, paper describes method abstract these normals parametric surface models. The works at two consecutive stages. Firstly, hierarchically grouped divide up into superpixels. These superpixels capture patches on that belong same surface. Then, they merged segment simple geometric Curvature model used planes, cylinders and/or spheres. This shows how, in higher levels abstraction scenes can be described using items their topological relationships.