Section: New Results

Temporally consistent 3D meshing from video data

Participants : Dobrina Boltcheva, Phuong Ho, Bruno Lévy.

This work is a part of the ANR Morpho project (Morpho ) which aims at combined analysis of human shapes and motions. In particular, the goal is to study how motions relate to human shapes or how shapes deform in typical motions. During this year, we addressed the first challenge which is building temporally consistent 3D meshes from silhouette images. We have already achieved a very fast meshing algorithm for each frame based on the Centroidal Voronoi Tessellation which has been previously developed in our team. Actually, we are investigating different ways for adding the temporal consistency within our optimisation framework. In particular, we are studding a strategy based on the optimal transport paradigm.