home   |<<   <--   -->   >>|

Discussion

I think its useful to know that 4x4x4 filters based on cubic polynomials are actually all you need to get decent curvature measurements,

and I think the bigggest challenge now is trying to measure the curvature of large features (like the skull in the images before), in the presence of noise, either from measurement, or from quantization.

I think the NPR work can be expanded, by playing with different styles of contours, different kinds of effects such as halos.

Finally, my implementation is really slow, but I'd love to learn about the extent to which its possible on current graphics hardware.