Implicit Surfaces Cluster-based Reconstruction from Scattered 3D Points Clouds

G. Monno and A.E. Uva (Italy)


Implicit Surfaces, Reconstruction, Clustering, Radial Basis Function


This paper presents a rather new method for object reconstruction from unstructured 3D points cloud using smooth implicit surfaces. The proposed method consists in generating a first reconstruction using hierarchical nearly planar “tiles” then we can construct an implicit surface interpolation using the centers and the normals of the tiles as constraints. This algorithm can generate multiresolution representations by iterating the hierarchical clustering step. The main innovative features of our reconstruction method are: • the practical possibility of using a very large point cloud for the implicit interpolation, still preserving low computational costs; • the lack of the necessity to obtain the normals directly from the scanner since those normals are evaluated in the clustering algorithm. The method has been successfully applied to both synthetic and real data and it is particularly suitable for reconstruction of human organs from medical data and soft-organic shapes in general.

Important Links:

Go Back