Fig 1.
System architecture.
Fig 2.
Two multi-camera acquisition platforms of different sizes.
1m×1m×1m(left) and 3m×3m×3m(right).
Fig 3.
Extended VbVH-based 3D reconstruction.
(a) Visual hull computation; (b) Surface corrected by depth map data; (c) Texture mapping.
Fig 4.
Compute the intersection point of the silhouette and the projection line.
Fig 5.
Geometrically-based deformation region (colored green) and physically-based deformation region (colored blue).
Fig 6.
Velocity-position correction for avoiding penetrations.
Fig 7.
Finding the completely cut elements.
Fig 8.
Generating the virtual nodes for the nodes of completely cut elements.
Fig 9.
Allocating the virtual nodes to generate new elements.
Fig 10.
Creating new generated elements.
Fig 11.
Multi-GPU implementation method.
Fig 12.
Parallel processing of voxel state sequence.
Fig 13.
Neighboring particle search method.
Table 1.
Experimental environment.
Fig 14.
The reconstructed results of bunny and hand.
Fig 15.
Comparing the reconstruction result with traditional VBVH method.
Fig 16.
The deformations of bunny modeled by the improved hybrid method.
Fig 17.
The reconstructed objects interact with the virtual deformable bunny.
Fig 18.
Cutting the virtual liver using the reconstructed scalpel.
Fig 19.
The virtual cutting simulation with bleeding effects.
Table 2.
Time performance comparison of 3D reconstruction methods(ms).
Fig 20.
Time performance of the improved hybrid deformation model.
Table 3.
Time performance comparison of texture mapping method(ms).
Table 4.
Time performance comparison of neighboring particle search methods(ms).
Table 5.
Time performance comparison of virtual-reality interactions(ms).