Accelerated, high-quality refraction computations for volume graphics

We present an efficient framework for the high-quality rendering of discretely sampled surface-based objects with refractive effects. This requires an accurate estimation of the refraction coefficients, paired with efficient and accurate surface detection, space traversal, and backdrop image sampling. Our framework achieves these goals, by employing a high-quality spline-based filter in conjunction with a novel filtered octree space decomposition with pre-classified cells that is carefully matched to the filter and voxel neighborhood characteristics. Here, we benefit greatly from the non-negativity of the B-Spline kernel. Finally, we describe an innovative scheme that achieves the high quality of pixel super-sampling on a flat backdrop plane without the overhead of tracing the actual rays across the refractive object.