Modeling

Point-Based Reconstruction

Participants: Patrick Reuter, Tamy Boubekeur, Christophe Schlick.

Recent 3D acquisition technologies provide a huge number of unorganized 3D points. Our goal is to develop new methods to reconstruct implicit surfaces from such large unorganized point sets. The driven idea of our approach is to divide the problem into three relatively independent steps: first, partition the global reconstruction domain into smaller overlapping local subdomains by using adaptive domain decomposition methods, second, solve the reconstruction problems in the local subdomains using some well-chosen function basis, and finally, blend the solutions together using the partition of unity method

We first investigated the use of radial basis functions with global support to achieve the second step. We thus proposed two methods using different space partitioning strategies: the first one[TRS04a]uses an adaptive decomposition based on a octree and blends the solution of all the leaf nodes together, while the second one uses a perfect binary tree and blends together at each nodes, the solutions of its two children. Both methods have a nice linear behavior for the required reconstruction time and memory usage, with the respect of the number of initial points.

More recently, we have investigated an alternative to radial basis functions during the local reconstruction step. Our new techniques is based on the "Enriched Reproducing Kernel Particle Approximation" (ERKPA) which was developed in the field of mechanics, during the late 90s, to solve some partial differential equations on mesh-less structures. The nice property of ERKPA is to allow the inclusion of sharp edges in the reconstructed implicit surface.

Point-Based Texturing

Participants: Patrick Reuter, Tamy Boubekeur, Christophe Schlick.

Our point-based reconstruction techniques described above, are not limited to the reconstruction of an implicit surface, but also apply to the reconstruction of continuous functions of any dimension. As a direct extension, we used the reconstruction methods to define a new class of procedural solid textures that can be reconstructed from the attributes of unorganized point sets.

We developed a new interactive environment for constructive texturing of surface of arbitrarily defined 3D objects[RSPS04]. A user can texture the surface by defining space partitions that are combined using constructive texturing, and by specifying attributes that are applied in the space partitions. In order to give an interactive feedback, a point-based multi-resolution representation of the surface is used not only for rendering, but also for the evaluation of the texture. We always keep a feedback to the initial geometric representation of the object (polygonal mesh, parametric or implicit surface, voxel arrays, etc) which means that the final textured object can be easily exported to standard graphics software that cannot directly handle discrete surface points (e.g. CAD systems, photorealistic rendering software, etc).

Adaptive Sampling of Implicit Surfaces

Participants: Florian Levet, Patrick Reuter and Christophe Schlick.

In this work [LHRS04,LHRS05], we propose a solution to adapt the differential point rendering technique to implicit surfaces. Differential point rendering was initially designed for parametric surfaces as a two-stage sampling process that strongly relies on an adjacency relationship for the samples which does not naturally exist for implicit surfaces. This fact made it particularly challenging to adapt the technique to implicit surfaces. To overcome this difficulty, we extended the particle sampling technique developed by Witkin and Heckbert in order to locally account for the principal directions of curvatures of the implicit surface. The final result of our process is a curvature driven anisotropic sampling where each sample "rules" a rectangular or elliptical surrounding domain and is locally oriented according to the directions of maximal and minimal curvatures. Like in the differential point rendering technique, these samples can then be efficiently rendered using a specific shader on a programmable GPU.

We put our sampling technique in the modeling section and not in the rendering one, because our final goal with anisotropic particles is not specifically to render the corresponding implicit surfaces, but rather to consider these particles as a set of control parameters that can be used to manipulate the surfaces, for instance, in a virtual sculpting environment.

Sketching Interface

Participants: Xavier Granier.

On mobile devices, the interaction is mainly done with a stylus. Thus, for the 3D modeling, the natural metaphor will be the Sketching for Modeling. Since it is difficult to perform precise 3D rotation with a stylus, we are developing currently a shape from shading approach for 3D modeling, in which the user will only have to place light direction and the resulting shading in order to create a 3D volume. Prelimirary results have been recently published [KGB05], and we are currently investigating the possibility of 3D model edition using similar techniques.

Creation of digital landscapes from maps

Participants: Joachim Pouderoux, Jean-Christophe Gonzato, Pascal Guitton.

The main goal of this subject is to create quasi-automatically a virtual environment from digital supports like scanned topographic maps, pictures. The map contains complex data to recover: DEM (Digital Elevation Model) from contour lines, environment objects from forest, houses, roads etc. The reconstruction of original DEM from contour lines consists in a combination of semi-automatic operations by using classical image analysis, neuronal networks and morphological mathematics: selection of contour lines [SG04], of skeleton, broken contours reconstruction, altitude affectation and finally interpolation of the DEM[PTGG04]. We developed a software named AutoMNT that load any topographic map and convert it in a classical DEM model [PGG03,PGGG04].

We are currently working on retrieving other essential information (road, forest, town, etc) and to integrate them in a complete Geographic Information System (GIS) database in order to view them in different environments: Reality Centers, standard PC, Web and PDA.

Scanned maps to Digital Elevation Models