Josh English wrote:

> The normal is an imaginary vector at each vertice of the triangle that
> declares which way is "up" to that point. By adjusting the "up" vectors at
> each point in a mesh of smooth triangles you can create the illusion that the
> bump surface of the triangles are joined into one smooth surface, you don't
> actually change the surface of the triangle, just it's appearence.

    I haven't looked at the details in nearly two decades but ... is it that one
creates a vector out of the vertex averaging the angle of the two sides and
pointing in the plane. Then two other triangles meet at the same vertex. The three
vectors are averaged and normalized and thus the mesh point angle. And then that
is averaged for midpoints with normal to the center of each triangle.

    And then how many times it is interated is how smooth it is.

    Is that about right?

    The rational rendering routine works backwards from the desired screen
resolution and determines the number of interations required so as to have less
than a pixels difference.

    That is either known or a Master's thesis in it.

    What I have not seen in any writeup is image resolution being a consideration
in rendering.

Post a reply to this message