Techniques for performance-based facial animation

M.Sanchez, "Techniques for performance-based, real-time facial animation", PhD, 2006. (Supervisor: Dr Steve Maddock). [pdf (author's copy)]

performance capture PhD Abstract: The purpose of this research project has been to construct a real-time Facial Animation system that reproduces a wide range of aspects of facial skin motion using information captured from actual performers. The large-scale deformation of facial skin is controlled by marker-based optical Motion Capture through the application of geometry-warping algorithms specially developed for this purpose. Smaller effects such as wrinkling and buckling are addressed through shading techniques applied onto the warped facial geometry, evaluating a model of fine-scale tissue behaviour that is also built using data retrieved from actual subjects. The synthetic vision framework required for capturing such data is also introduced in this thesis. Additional aspects regarding the analysis and reusability of Motion Capture data are also considered, enabling the system to apply the data collected from a specific individual to different physiognomies. The real-time nature of this animation system is reinforced by implementing some of its time-critical components in dedicated hardware (programmable graphics cards), and preprocessing the facial geometry to make better use of such hardware. Finally, recommendations are made for future work of this nature.

Publications

Downloads

Invited talks