Closing Gaps by Clustering Unseen Directions

• Venue: SMI '04 Proceedings of the Shape Modeling International 2004
• Year: 2004

Although in recent years the 3D-scanning field hasreached a good level of maturity, it is still far from beingperceived by common users as a 3D-photography approach,as simple as standard photography is. The mainreason for that is that obtaining good 3D models withouthuman intervention is still very hard. In particular,two problems remain open: automatic registration of singleshots and planning of the acquisition session. In thispaper we address the second issue and propose a solutionto improve the coverage of automatically acquired objects.Rather than searching for the next-best-view in orderto minimise the number of acquisitions, we propose a simpleand easy-to-implement algorithm limiting our scope toclosing gaps (i.e. filling unsampled regions) in roughly acquiredmodels. The idea is very simple: detect holes in thecurrent model and cluster their estimated normals in orderto determine new views. Some results are shown tosupport our approach.