Abstract
Creation of 3D graphical content becomes ever harder, as both display capabilities and the demand for complex 3D content increase. In this thesis, we present a method of using densely scanned surface data from physical objects in interactive animation systems. By using a layered approach, incorporating skeletal animation and displacement mapping, we can realistically animate complex datasets with a minimum of manual intervention. We propose a method using three layers; firstly, an articulated skeleton layer provides simple motion control of the object. Secondly, a low-polygon control layer, based on the scanned surface, is mapped to this skeleton, and animated using a novel geometric skeletal animation method. Finally, the densely sampled surface mesh is mapped to this control layer using a normal volume mapping, forming the detail layer of the system. This mapping allows animation of the dense mesh data based on deformation of the control layer beneath. The complete layered animation chain allows an animator to perform interactive animation using the control layer, the results of which can then be used to automatically animate a highly detailed surface for final rendering. We also propose an extension to this method, in which the detail layer is replaced by a displacement map defined over the control layer. This enables dynamic level of detail rendering, allowing realtime rendering of the dense data, or an approximation thereof. This representation also supports such applications as simple surface editing and compression of surface data. We describe a novel displacement map creation technique based on normal volume mapping, and analyse the performance and accuracy of this method.