CageR: From 3D performance capture to cage-based representation

Modern performance capture systems [de Aguiar et al. 2008] provide high resolution 3D mesh sequences which are becoming critical components for today's special effects. Unfortunately, such raw sequences have a large memory footprint and are difficult to edit. We propose CageR, a framework based on spatial deformation with cages to construct automatically a compact and editable high level representation of these raw sequences, resulting in high compression factors and allowing easier post processing. In particular, we formulate an automatic cage fitting algorithm embedding a new relaxation strategy based on Maximum Volume and a new regularization method based on sub-spectral analysis. As a result, we use the CageR representation in various applications, including compression, motion transfer and shape space modeling.