"Bald Eagle" <cre### [at] netscapenet> wrote:
> "Chris R" <car### [at] comcastnet> wrote:
>
> > I'd be interested when you find your method of doing this as well.
>
> I'll keep digging in the archive.
> So. Many. Files.
>
> The general gist of the method was like doing a Laplacian - rather than
> evaluating the isosurface at <x, y, z>, Something like an inside test was done
> across the 9 or 27 "voxels" surrounding and including the point.
> The sum of this would then _gradually_ drop off, as more and more sample points
> were zero, and only a few were inside the object.
>
> I guess it's also like dithering, and edge-finding with a convolution kernel.
>
> I could swear that I mapped / graphed this all out and got it to work very
> nicely - I just have to find it.
>
> - BW

That was actually the first approach I thought about; it's very similar to what
Inigo Quilez proposes for doing repeating patterns of SDFs.  I had trouble
visualizing how much variance it would introduce in the edges while introducing
27 calls to the pigment function at each point.  Did you find you needed to
expand the box beyond the neighboring points, perhaps to 2 points out, which
would be 125 calls to the pigment function to get a good gradient for shading?

-- Chris R

Post a reply to this message