Variations of the Splitting Box Scheme forAdaptive Generation of Contour

A widespread approach to generating iso-surfaces or contour surfaces in volume data is the so-called marching-cubes algorithm. This algorithm, however, has the disadvantage that the number of polygonal chains generated is considerable. Several attempts to reduce the size of output were undertaken in the past. For instance, we have suggested the splitting box approach. An important advantage against other adaptive solutions is that the problem of cracks inherent to adaptive approximations is solved in a way that neighboring boxes can be processed independently without storing additional context information. In this contribution, a new algorithm is described. It basically follows the splitting box approach, but now bilinear interpolation on grid faces and trilinear interpolation on grid cubes are used. It turns out that the new approach is conceptionally easier to describe and to implement than the original splitting box algorithm. Experimental investigations on several data sets however showed that the data reduction is inferior compared with the splitting box algorithm. From an analysis of the causes an improvement of both algorithms is derived. Further investigations show that the results of the splitting box algorithm are almost optimal. Variations of the Splitting Box Scheme for Adaptive Generation of Contour Surfaces in Volume Data