Data reduction by applying an image-based modeling and rendering technique to CG models

We applied an image-based modeling and rendering technique to reduce the data size of a CG model to be obtained as a visualization result. To date, this technique has been applied to the reconstruction of 3D graphical models from real objects; the size of the models can be reduced effectively because both color information and geometric information can be reduced. We applied a silhouette-and-voxel method that does not require design data for geometric simplification. Using our system, we simplified two models, one of which, involving medical data, was reduced by about 85% in file size. An accompanying subjective quality test showed that our approach maintains approximately the same visual quality as the geometric simplification method traditionally used.