Non-rigid Face Modelling Using Shape Priors

Non-rigid 3D shape recovery is an inherently ambiguous problem. Given a specific rigid motion, different non-rigid shapes can be found that fit the measurements. To solve this ambiguity prior knowledge on the shape and motion should be used to constrain the solution. This paper is based on the observation that often not all the points on a moving and deforming surface such as a human face are undergoing non-rigid motion. Some of the points are frequently on rigid parts of the structure – for instance the nose – while others lie on deformable areas. First we develop a segmentation algorithm to separate rigid and non-rigid motion. Once this segmentation is available, the rigid points can be used to estimate the overall rigid motion and to constrain the underlying mean shape. We propose two reconstruction algorithms and show that improved reconstructions can be obtained when the priors on the shape are used on synthetic and real data.

• Type: Conference
• Year: 2005
• Where: AMFG
• Authors: Alessio Del Bue, Xavier Lladó, Lourdes de Agapito
• Keywords: AMFG 2005, Biometrics, Non-rigid 3D Shape, Non-rigid Motion, Rigid Motion