SIGGRAPH Course 2014 — Skinning: Real-time Shape Deformationesh Animation Decomposition and Compression

With the development of advanced computer vision and tracking techniques, deformed mesh sequences can be soundly reconstructed by markerless performance capture [De Aguiar et al. 2008b; Vlasic et al. 2008; Vlasic et al. 2009; Stoll et al. 2010], or by motion capture with dense markers [Park and Hodgins 2006; De Aguiar et al. 2007]. With the maturity of these high-quality capture techniques, researchers in computer graphics community have also explored the concept of using example poses (i.e., a sequence of deformed mesh frames) for character skinning, rigging and animation, which has become increasingly practical and useful in entertainment practice. For example, proxy bones (or called bones for simplicity) and corresponding skinning weights can be automatically extracted from a set of example poses [James and Twigg 2005; Kavan et al. 2010; Le and Deng 2012], and the resulting bones and skinning weights can be potentially used for many applications including mesh animation compression, hardware-accelerated skinning animation, collision detection, animation editing, and so on [James and Twigg 2005; Kavan et al. 2010]. Furthermore, different from a set of disconnected proxy bones, a skeletal rigging model can also be automatically extracted from example poses [Schaefer and Yuksel 2007; Hasler et al. 2010; Le and Deng 2014]. Since the skeleton extracted from example poses is compatible with game engines and popular animation software such as Maya and Blender, it can be directly used for various animation editing, compression, and rendering applications, which could help to substantially reduce production cost in industry practice.