Visualization of Large Rle-encoded Voxel Volumes

We present a method for visualizing large voxel volumes based on optimized ray-casting. Other than conventional methods casting a ray for each pixel on the screen, our method only casts one ray per column and then traverses the voxel volume in a front to back manner. This can be done efficiently as our data is encoded by run-length-encoding (RLE), reducing the overall cost for the traversal. To exploit frame-to-frame coherency and to make the visualized scene rotation invariant, we are storing the rendered result temporarily in a cube-map. As the cube-map might have any orientation when finally rendered as cube around the view-point, 6 DOF are achieved.