Stage Master 2 en segmentation et édition de maillage

19 Février 2023

---

Catégorie : Stagiaire

---

Supervision and contact

Pr. Bertrand Kerautret, LIRIS (bertrand.kerautret@univ-lyon2.fr)

Localization

The candidate will be attached to the LIRIS laboratory located at Campus Porte de Alpes (Bron).

Motivations

In the context of shape acquisition and modeling, the identification and separation of geometric structures can be of key importance to simplify, edit or improve the quality of a mesh. These research axis have concrete applications in video games where 3D shapes can be modeled from scans and then edited and re-meshed and eventually animated from a skeleton. In parallel, recently the acquisition methods based on photogrammetry have been democratized and now allow to produce 3D reconstructions from a number of photos without manual editing or complex parameters.

The other motivation of this topic is to continue the work based on the accumulation of normals [1,2] which introduced a method of centerline extraction able to handle partial meshes without geometric or topological constraints. This method relies on a simple idea of ray tracing that accumulate score in the digital space. The demonstration of the base method is illustrated on the following online demonstration and the source code is available on GitHub .

In this internship the idea is to continue exploring the concepts of this method through segmentation, smoothing, editing and reconstruction objectives. The main idea is to exploit the simplicity and the implementation of the method based on accumulation to adapt it to the exploitation in a concrete product as a plug-in for Blender.

Subject

In this internship, we propose to explore different perspectives of the accumulation algorithm through the following directions:

• Interactive selection of tubular structures.
The accumulation-based centerline extraction algorithm also retrieves an estimate of the thickness that is directly given through the maximum value of the accumulation value computed along the ray. A possible method could be to introduce a growing region based algorithm defined from the selection of a face and to exploit the discrete accumulation space This space would allow to constrain the region growth by integrating the geometric information accumulated in the voxels.
• Split of tubular parts.
Thanks to the previous point, it will be possible to discriminate tubular parts even if they are physically connected to the rest of the mesh. For example, a goal would be to propose a tool allowing to separate and detach the arm from the rest of the body (like the right arm represented on the mesh of the image (a) of the figure. Following the progress of this point, it could be interesting to propose the implementation of this algorithm in a plug-in for the 3D modeling software Blender.
• Smoothing and surface reconstructions
Another point of interest could be to study the exploitation of the accumulation algorithm to define a surface smoothing or a partial reconstruction of tubular sections.

[1] B. Kerautret, A. Krähenbühl, I. Debled-Rennesson, and J.-O. Lachaud. Centerline detection on partial mesh scans by confidence vote in accumulation map. In ICPR, pages 1376–1381, 2016. [2] Bertrand Kerautret, Adrien Krähenbühl, Isabelle Debled-Rennesson, and Jacques-Olivier Lachaud. On the Implementation of Centerline Extraction based on Confidence Vote in Accumulation Map. In Pro- ceedings of the First Workshop of Reproducible Research in Pat. Rec., volume 10214, pages 109–123. Springer, 2016.

Awaited skills

Programming (C++ and/or Python), image analysis and processing. A experience in image processing, in the use of C++ libraries (e.g. DGtal libraries (e.g. DGtal) or graphical user interface programming (QT) or plug-in programming would be a plus in the application.

Applications

To apply, you can email your CV and latest transcripts to the following address bertrand.kerautret@univ-lyon2.fr

Potential extension

The basis of this master subject could be declined into a thesis subject that could be linked to other types of applications in the field of application with the extraction of blood vessels or the analysis of roots in collaboration with the INRAE.