Kari Kivisalo wrote:
> The method of calculating the normal vector from the scalar
> field may not be precise. Averaging these vectors would leave
> a small residue or error. This is purely a guess based on the
> existence of accuracy keyword for normals.

Increasing the accuracy (that is, making the number smaller) actually 
makes the problem worse. Decreasing it helps to smooth out the results, 
but only inasmuch as decreased accuracy smoothes out *any* normal.

The only sensible explanation I can think of would be if negative normal 
values somehow transform the normal, thereby causing two unrelated sets 
of numbers to be averaged.. but I can't imagine why this would be the 
case. (Besides, translating one of the normals by 100 units creates an 
entirely different effect.)

-Xplo

Post a reply to this message