Multiresolution volume data structure for shape modeling

Abstract

Due to the rapid progress of computer hardware, handling large volume data is getting easier. Although a number of methods have been developed for rendering volumetric datasets, very few have been proposed for shape modeling of such datasets. The paper describes a multiresolution structure for shape modeling with volume data, and the voxel interaction method capable of modeling shapes at arbitrary resolution level. In the voxel interaction method, each voxel acts in a similar way to a cellular automaton. The shape of an object is represented by the internal variables in the voxels. Unlike cellular automata, the interacting voxels can change the speed of the information propagation by selecting an appropriate resolution in the hierarchical structure, which plays an important role in modeling the rigidity of objects in our method. For the acceleration of rendering for 2D image generation, the sampling ratio along the eye-ray in the volume space is determined according to the resolution of the volume data. Experimental results include the examples of shape deformation of human CT data.