Tutorials
T2: Geometric Registration for Deformable Shapes
Monday, May 3. 09:00–17:00 in Conference room 1 (full day tutorial)Presenter: Will Chang, QiXing Huang, Hao Li, Niloy Mitra, Mark Pauly, Michael Wand
Abstract
Capturing the shape of a deforming 3D object remains an important and challenging topic. The key problem is to establish correspondences across time-varying partial scans obtained from multiple view directions, and bring the scans into alignment. Not only is an automatic algorithm for tracking and finding correspondences useful for shape reconstruction, but also it is the foundation of further 3D processing. Applications range from statistical analysis and modeling of dynamic objects, to post-production of special effects in movies and entertainment.
This tutorial overviews recent exciting and rapidly evolving research on registration and correspondence finding across multiple deforming shapes. We address the basic techniques of three main topic areas:
- Pair-wise local optimization methods
- Pair-wise global optimization methods
- Multi-scan correspondence and temporal sequence reconstruction.
The tutorial gives an introduction to deformable shape matching for attendees with a basic geometry processing background as well as providing an overview of the current state-of-the-art.