<--back--   home   --next-->

First- the hue-balls. The main goal here was to find a way of assigning color to the tensor that did not rely on simplifying the diffusion tensor down to just its principal eigenvector. And my reason for this was to avoid the sort of discontinuity that you can get in cases of planar anistropy. Here's a sequence of slowly varying ellipses with mostly planar anisotropy, with red and green lines showing the principal and secondary eigenvectors, respectively. Notice that between here and here the direction of the principal eigenvector shifts 90 degrees, even though these two tensors are really very similar. I think its important for a tensor coloring scheme to give similar colors to similar tensors, such as these two in the middle. Along these lines, I'm interested in using color as a sort of visual glue- so that coherent structures in the tensor field get the same basic color, and so that disjoint structures get dissimilar colors.