Deformable objects may be needed in many computer graphics and virtual reality applications. They are particularly useful in modeling clothes, facial expression, human and animal characters. A common method to render these objects is by tessellation. However, the tessellation process is computationally very expensive. As an object is deforming, we need to retessellate the surfaces in every frame. This computational burden poses a significant challenge to the real-time rendering of deformable objects. Consequently, deformable objects are seldom supported in existing virtual reality systems.

In this project, we are developing techniques to support real-time rendering of deformable objects. Our earlier work focused on objects represented by NURBS surfaces and we have developed a very efficient method for rendering deforming NURBS surfaces. The basic idea of our method is that we maintain two data structures of the deforming surface, the surface model and a polygon model representing the surface model. As the surface deforms, the polygon model is not regenerated through polygonization. Instead, it is incrementally updated to represent the deforming surface. Two techniques are fundamental to our method: incremental polygon updating and resolution refinement.

• Incremental polygon updating: It is based on the incremental property of NURBS surfaces. We first precompute a polygon model to represent each NURBS surface and a set of coefficients for each control point of the surface. During run-time, we compute the displacement of each affected control point as the surface deforms. Instead of retessellating the surface again, we update the polygon model by the displacement of each control point and the corresponding set of coefficient.
• Resolution refinement: When a surface deforms, its curvature is also changed. If the curvature is increased or decreased by a large amount during the deformation process, the resolution of the corresponding polygon model may become too coarse or higher than necessary to represent the deformed surface, respectively. To overcome this problem, we have developed a method to refine the resolution of the polygon model and to generate the corresponding deformation coefficients incrementally according to the change in local curvature of the surface.

Click here for a gzip'ed demo video (5MB in size) of interactive editing a human face model.

Click here for a demo video (3.7MB in size) on detecting inter-collisions of two NURBS surfaces.

Click here for a demo video (3.2MB in size) on detecting self-collisions of a NURBS surface.

Click here for a demo video (5.3MB in size) on trimming and deformation of a teapot model.

Click here for a demo video (6.6MB in size) on deformation of a trimmed head model.

Publications:

• Gary Cheung, Rynson Lau, and Frederick Li, "Incremental Rendering of Deformable Trimmed NURBS Surfaces," Proceedings of ACM VRST, pp. 48-55, Oct. 2003.
• Rynson Lau, Oliver Chan, Mo Luk, and Frederick Li, "A Collision Detection Framework for Deformable Objects," Proceedings of ACM VRST, pp. 113-120, Nov. 2002.
• Frederick Li and Rynson Lau, "Incremental Polygonization of Deforming NURBS Surfaces," Journal of Graphics Tools, A.K. Peters, 4(4), pp. 37-50, 1999.
• Frederick Li and Rynson Lau, "Real-Time Rendering of Deformable Parametric Free-Form Surfaces," Proceedings of ACM VRST, pp. 131-138, December 1999.
• Frederick Li, Rynson Lau, and Mark Green, " Interactive Rendering of Deforming NURBS Surfaces," Computer Graphics Forum (special issue on Eurographics '97), 16(3), pp. 47-56, September 1997.

Last updated on 21st August, 2003.