This paper presents a new method that combines the use of medial axis and implicit surfaces in order to reconstruct a 3D solid from an unstructured set ofpoints sampled on its surface. The proposed method uses implicit iso-surfaces generated by ponctual skeletons, that are a particularly compact way of defining a smooth free-form solid. As in a previous work the principle is the minimization of an energy that represents the distance between the set of points and the iso-surface. But our method provides an efficient and robust initialization, through the computation of the medial axis of the set of points that allows to reconstruct objects of complex topology, such as objects with holes and branchings or composed of several disconnected components. Moreover, instead of subdividing existing skeleton points to refine the reconstructed shape, we propose a heuristic, based on the notion of ``local skeleton energy'',to select iteratively new useful skeleton points from the pre-computed medial axis data. This heuristic drastically speeds up the reconstruction process. The method can be implemented in a fully automatic way and has been successfully applied to both synthetic and real data.
| Citation
info
|
@inproceedings{BTG95,
author = "Eric Bittar and Nicolas Tsingos and Marie-Paule Gascuel", title = "Automatic reconstruction of unstructured 3D data : Combining a medial axis and implicit surfaces", booktitle = "Proceedings of Eurographics'95,Maastricht, The Netherlands", year="1995" } |