Anatomically based rigging using layered free form deformation

Abstract

Rigging a human body model is an active research area in Computer Animation. When a body moves the shape of the body undergoes non-rigid articulated motion. The skin is the surface of a human body and defines a shape of the body. In Computer Graphics, it is a common practice to model only the surface of a body and to apply the deformation of the shape directly onto the surface. However, underneath the skin, there are muscles, fatty layers and bones where their positions and shape affect the final appearance or the shape of skin. There have been anatomy-based modeling approaches to mimic a real human body structure, but the muscle models are either too simple to exhibit a realistic properties of the muscles or unnecessarily too complex for computer animation. Moreover, the skinning is rather arbitrary and the shape control is not intuitive. In this thesis, we propose a new representation of a human body for an intuitive rigging based on $\It{layered}$ free-form deformation. We start with NURBS-based free-form deformation (NFFD) in order to allow natural shape of skin and muscles. For modeling muscles, we extend the NFFD to incorporate some physical properties such as volume preservation. For a skin model, the initial shape is represented by NURBS-based FFD, but its deformation is based on a layered free-form deformation reflecting the anatomical structures of a limb, where the muscles are attached to a bone and the muscle layer is underneath the skin affecting the final shape of the skin, based on the articulated motion of the bone. We define various types of muscle using our extended $\It{constraint}$-NFFD, and represent the human body using $\It{layered}$ FFD. We demonstrate that a motion of a joint creates a realistic deformation of the skin in real-time.