Surface Reconstruction from 3D Scanning

This research area consists in general methods for automatic reconstruction of accurate, concise, piecewise smooth surfaces from unorganized 3D points tipically acquired by a 3D scanner.

Our research is motivated by numerous applications: reverse engineering, design, inspection, planning of medical treatment, custom fitting, textured objects, populating virtual worlds, display objects on internet and special effect for films. About this latter we remember that Computer Graphics is increasingly used in films. Special effects that would be otherwise impossible, infeasible, or just expensive can be digitally combined with video sequences. The extra character, objects, or background tend to look more realistic if they are scanned from real counterparts than if they were completely generated by a computer.

It is possible to develop algorithm for surface reconstruction without having actual range scanner. So in literature we find a lot of proposal starting by sampling points from a variety of existing surface models and this was useful to compare results with known references.

Recently we have bought a touch-probe 3D scanner (see figures) and we have realized that the most difficult problem is not to reconstruct a surface from a set of 3D points, but to obtain a valid set of 3D points from a set of range images and that this activity is strictly scanner type dependent.

PIX 30 Acquisition process

In the following figure we show a Budda which is acquired with our scanner in four step o range images, than these are merge and filtered to obtain a first set of 3D valid points; note that the different point colours mean they are taken from different range images.

Budda points

About the surface reconstruction methods we are exploring both traditional NURBS fitting tecniques and more powerful Subdivision and Radial Basis Function surfaces interpolating and approximating.

In the following web site you can view some models that were scanned and recostructed during our research project.
Some reconstructed models