Graphics: Visual and Interactive Computing
Volume Morphing and Applications
Please also visit the Swansea volume graphics gallery. |
Geometric deformation is an important aspect of computer animation. With traditional surface representations, the control of the deformation is usually achieved by moving the vertices of a polygonal mesh, or the control points of a parametric surface. A range of methods have been proposed for specifying and controlling surface deformation. During the past decade, we also witnessed the rapid development of image distortion techniques, which played an important role in creating stunning visual effects in the entertainment industry. Because of the extra dimension of information in volume representations, the deformation of a volume object is more difficult to achieve. However, research has shown that it is technically feasible, and potentially more desirable than surface- and image-based deformation in many applications. Volume deformation defines a geometric transformation from one volume object to another under the influence of two control datasets, each of which usually consists of a set of geometric objects representing the main features of a volume. Given a starting volume Va and a finishing volume Vb, volume morphing is a process that generates a sequence of inbetween volumes V1, V2, ..., Vn which represent a smooth transformation from Va to Vb. Let Ca and Cb be the control datasets associated with Va and Vb respectively. For each inbetween volume Vi, an inbetween control dataset Ci is first obtained as an interpolation of Ca and Cb. Ci is then used, in conjunction with Ca and Cb respectively, to deform Va and Vb, resulting in two distorted volumes Via and Vib. The inbetween volume Vi is then obtained as the interpolation of Via and Vib. According to the use of control datasets, approaches to volume morphing can be classified into three categories, namely cross dissolving, mesh warping and field morphing. At Swansea, we have studied a range of volume deformation and morphing algorithms. In particular, we have developed a disc field morphing algorithm which produces good quality results at a reasonable speed without demanding complex control information. Its flexibility and controllability are apparent, and can be applied to the general form of volume data. We have also applied the concept of volume deformation to forensic science for facial reconstructing. The ultimate aim of any facial reconstruction technique is to produce a likeness of the face from skeletalised, burnt, badly mutilated or decomposed remains so that it bears sufficient resemblance to the individual prior to death. Traditional sculpting methods rely on a combination of the ability, anatomical and anthropological knowledge of the artist and numerous subjective interpretations on the form of the face to produce a reconstruction of an unknown skull. Computer-aided reconstruction techniques are both quicker and more flexible but rely heavily on the size and quality of the facial database from which the reconstructions are created. With volume-based deformation, the problem of facial reconstruction is approached from a perspective where the facial soft tissues are considered as a single unit. It manipulates a reference volume within the anatomical confines of the sample from which it is derived. This allows for realistic transitions to be made to the soft tissues of the face in keeping with the differences in the form of the target skull on which the reconstruction is based. Simon Michael's undergraduate project, which was a feasibility study of volume-based facial reconstruction, received 1995 WDA Technology Prize. He continued this work through his PhD programme, which was completed in 1999. Main References
|
