Create Virtural Human:


We are developing a fast segmentation freamework providing a general solution to create virtual human bodies. This segmentation work has been done within the CRaIVE project which is building VE's for training in vascular interventional radiological procedures. Figure 1 and 2 show the results of blood vessel segmentation from MRA images.


 

Figure 1 Segmentation of blood vessel and skin (right) from the MRA image of the lower part of legs (left).


 

Figure 2 Blood vessel segmentation (right) from the MRA image of the upper part of body (left).



3D Metamorphosis:


In addition to the advantage of the compact 3D shape representation, this project provides an easy solution of establishing dense correspondences among objects. One of the applications is 3D metamorphosis. A series of intermediate objects can be automatically created by interpolating corresponding points from the source object to the target object, as shown in Figure 3.


Figure 3 Automatically created morphing sequence.


In addition to the the spatial correspondences, in Figure 4, we demonstrate that the similiar technique can be applied to build correspondences between texture mappings.


Figure 4 3D metamorphosis with automatically created texture mapping.



Direct Manipulation of Deformation:


For interactive applications such as video games and computer-aided design, real time performance and the ease of shape editing are of great concern. The shape representation developed in this project allows direct manipulation of free-form deformation, as shown in Figure 5.


Figure 5 Changing facial attributes (nose).



3D Synthesis:


Since shape descriptors proposed in this project are the only factor to distinguish the shape differencebetween objects, the goal of creating an “average” object can be achieved by simply averaging the shape descriptors. Figure 6 demonstrated a synthetic face (in the centre) by averaging 12 face models reconstructed from realistic human face data collected from a 3D scanner, which are shown around the average face.


Figure 6 The synthetic face by averaging 12 reconstructed face models.