Two-level volume rendering

摘要

Presents a two-level approach for volume rendering, which allowsnfor selectively using different rendering techniques for differentnsubsets of a 3D data set. Different structures within the data set arenrendered locally on an object-by-object basis by either direct volumenrendering (DVR), maximum-intensity projection (MIP), surface rendering,nvalue integration (X-ray-like images) or non-photorealistic renderingn(NPR). All the results of subsequent object renderings are combinednglobally in a merging step (usually compositing in our case). Thisnallows us to selectively choose the most suitable technique forndepicting each object within the data while keeping the amount ofninformation contained in the image at a reasonable level. This isnespecially useful when inner structures should be visualized togethernwith semi-transparent outer parts, similar to the focus+context approachnknown from information visualization. We also present an implementationnof our approach which allows us to explore volumetric data usingntwo-level rendering at interactive frame rates