Parallel construction and isosurface extraction of recursive tree structures

The visualization of volumetric datasets is usually limited by the amount of memory and processing power of computer systems. Several multiresolution methods have been developed in order to adapt the necessary work; recursive spatial tree structures, such as octrees, are among the most popular. The exploration of a dataset frequently requires a change of parameters, such as color table entries or isovalues. Therefore, the costly update of an octree becomes necessary. To overcome this drawback, we propose the parallel construction of octrees to improve their suitability for interactive volume visualization. Based on the thread model of the shared-memory paradigm, we developed a scheme for a balanced parallel construction. We apply this new scheme to generate isosurfaces in parallel, using the Marching Cubes algorithm