<--back--   home   --next-->

So, our paper explored using volume rendering for this problem, since the spirit of volume rendering is that you can get this more wholistic view of the entire volume. The volume rendering method we used was very simple ray-casting: no perspective, no global illumination, one ray per pixel, and regular sampling along the ray. At each sample point along the ray, we sample the tensor field by tri-linearly interpolating it component-wise. From the interpolated tensor we calculate color and opacity, and these values are composited for all the tensors along the ray ray to produce the pixel color. Now really, the only interesting question in all of this is, "how do you come up with color and opacity for a tensor?". Our paper explores one possible way of answering that question.