Shape Continuity-Driven 3D Surface Reconstruction from Parallel Cross Sections

3D shape reconstruction from parallel cross sections is central in medical imaging, computer-aided design, computer graphics, and other fields. Existing methods, e.g., NUAGES, are based on the identification of 2D Delone–Voronoi and medial axis complexes in levels i and i+1, which are used to construct a set of tetrahedra that fill the 3D space between the cross sections. To improve this process, our method, named NUAGES+, directly uses 3D Delone tetrahedra to materialize the evolution between consecutive cross sections by benefiting from a denser sampling of the original data. This combination produces more robust results, stemming from better shape ratios of the Delone tetrahedra. In addition, we use a 2D shape similarity filter that eliminates improbable solid connections, which originate in geometric overlaps between consecutive cross sections. Future work is needed in the algorithmic support for synthesis of 0-, and 2-handles to materialize the topological changes between consecutive sections.