Master2: Time-coherent meshing for 3D human motion acquisition

We are looking for a highly motivated Master student to work on the subject of time-coherent meshing for 3D human motion acquisition.

 


Centroidal Voronoi Tessellation [LLW12].

Context

Multi-view stereovision is an efficient way of capturing 3D moving objects. It is the foundation of many visual effects used in movies, such as the « bullet-time » effect in the « Matrix » series. This research topic is part of a cooperative MORPHO project, that uses multi-view stereovision to more accurately capture the motion of humen. MORPHO combines the expertise of researchers in biomechanics, computer vision and geometry processing, and offers an interesting opportunity of using the cutting-edge multi-view acquisition setup available in Grenoble (Grimage and/or 4DViews).

Objectives

The goal of the project is to define a new algorithm to optimize the representation of multi-view stereo data. Namely, the goal is to obtain a mesh that is time-coherent, that is to say, that uses the same discretization over time. Such a representation is more compact in memory and avoids some unwanted flickering effects. It also allows elaborate post-treatments, editing, indexing, recognizing sub-sequences, analyzing motion.

The algorithm is based on the notion of Centroidal Voronoi Tesselation [DW05,LLW12], used to optimize the sampling of an arbitrary domain. The objectives of the internship are:

  1. understand Centroidal Voronoi Tesselations;
  2. develop an algorithm that computes a Centroidal Voronoi Tesselation for multi-view stereo data (visual hull [Lau94]);
  3. add a regularization term to ensure time coherence.

Keywords
Computer vision, computer graphics, geometry processing, computational geometry, Voronoi diagram.

Student profile

  • Master student, preferably in computer science or applied mathematics.
  • Creative and highly motivated.
  • Solid programming skills; the project involves programming in C++, Matlab.
  • Solid mathematics knowledge (especially linear algebra, numerical analysis and geometry).
  • Fluent English or French spoken, and fluent written English.
  • Prior knowledge in the areas of computer vision, computational geometry, computer graphics and/or signal processing is a plus.

Duration: 5 to 7 months.

Start date: February 2013.

Location: Inria Nancy and/or Grenoble, France.

How to apply

Please send applications through this page:
http://morpheo.inrialpes.fr/job-applications/

  • a complete CV
  • graduation marks, rankings
  • the name and email address of references, if relevant

Advisors

Bruno Lévy, ALICE team, LORIA & Inria
Dobrina Boltcheva, ALICE team, LORIA & Inria
Franck Hétroy, Morpheo team, LJK & Inria
E-mail : bruno.levy@inria.fr
Tel.: 03 83 59 30 70
E-mail : dobrina.boltcheva@inria.fr
Tel.: 03 83 59 30 00
E-mail : franck.hetroy@inria.fr
Tel.: 04 76 61 55 04

References

  • [DW05] Q. Du and D. Wang: Anisotropic Centroidal Voronoi Tessellations and Their Applications. SIAM J. Scientific Computing 26(3): 737-761 (2005).
  • [Lau94] A. Laurentini. The visual hull concept for silhouette-based image understanding. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1994.
  • [LLW12] L. Lu, B. Lévy and W. Wang. Centroidal Voronoi Tessellations for Line Segments and Graphs. Computer Graphics Forum (Eurographics proceedings), 2012.