
But now we can do that with curvature based transfer functions.We vary the world-space contour thickness, that range of surface normals that we color differently, to create the appearance, in image-space, of a constant thickness contour.
This is possible with two pieces of information, the curvature along the view direction, which is easy to extract from the geometry tensor, and the desired image-space contour thickness, T, which is the one parameter the user controls.
And you do a little bit of trigonometry, and you find that you still express the contour in terms of v-dot-n, but now this side varies with the view direction curvature. We implement these contous with a 2-D lookup table, indexed by v dot n, and the curvature.