EN FR
EN FR


Section: New Results

New results

This year, we obtained new results in fabrication, in geometry processing and in multi-view reconstruction.

We investigated software solutions for printing with low cost (filament) 3D printers. We proposed a solution to automatically define temporary structures that will supports the object during its creation [9] . We also strongly reduce the artefacts that are produced by multi-material printing [17] . These works allow to better understand the physics of these printers, and to come up with efficient software solutions to common drawbacks of this technology. Other contributions in fabrication are more related to the design of the printable objects, that is developed in our software IceSL. To achieve real-time rendering of CSG models, we developed a new GPU approach for single pass A-Buffer [23] . This technique is also a simple solution to handle complex rendering problems such as transparency. We also proposed [11] an efficient method for performing dilatation and erosion directly on the same representation of volume by sequence of dilatation and erosions on segments.

In geometry processing, we proposed an algorithm to compute the intersection of Voronoi cells and a simplicial complex [25] . This algorithm is fast in dimension up to 10D because it doesn't require to explicit the Voronoi diagram. It comes with exact predicates and symbolic perturbation to ensure its robustness. We have also developed an algorithm [13] able to trace streamlines on triangulated surfaces in such a way that two such streamlines cannot cross or merge. This property seems obvious in the continuous case, but was very difficult to enforce with the discrete representations (triangulated surface, and floating points) manipulated by the computer. We did also revisit the Optimal Delaunay triangulation in the case of graded mesh generation [14] , and we adapted our remeshing methods to Geologic applications [27] .

We obtained some new results in multi-view reconstruction: a new method that expands a limited set of correspondences towards a quasi-dense map across two views [15] , and an improvement of variational multi-view reconstruction obtained thanks to a simple characterization of geometric deformations [16] .