0
\$\begingroup\$

Generaly when I think of marching cubes LOD mesh differences, I think of something like:

enter image description here

But the reallity is:

enter image description here

Theoretically the edges (right, left in picture) are same for low and high resolutions sides and should create equaly positioned vertices, but for some reason they differ slightly, what creates lot's of problems for stiching the cracks.

enter image description here

For some reason the vertex X calculated from blue line interpolation (lowe-res) and purple (hi-res) differs.

And I wonder if this is fixable? I checked and it's not interpolation issue, but has something todo with SDF (densities values) and especially vivid with round stuff like Sphere.

\$\endgroup\$
4
  • \$\begingroup\$ The Transvoxel algorithm provides one way to seamlessly bridge resolution changes in marching cubes meshes with LOD. \$\endgroup\$
    – DMGregory
    Jul 11, 2021 at 12:14
  • \$\begingroup\$ Yes I know, but that is not my question. I want to understand the cause of this differences so I can work on my solution. \$\endgroup\$ Jul 11, 2021 at 16:00
  • \$\begingroup\$ Can you show us how you're computing the vertex positions from your SDF? That could help isolate the cause. \$\endgroup\$
    – DMGregory
    Jul 11, 2021 at 16:08
  • \$\begingroup\$ I don't think so. It's lot of code in summary. Also I do not think it's some bug in my implementation (something special wrong) and it's more like common thing (voxelterrain.com - can see here also corner diff.) for marching cubes + LOD. I was just hoping that it's addressed somewhere some have in some paper or somebody personal experience that I have overseen. \$\endgroup\$ Jul 11, 2021 at 16:34

0

You must log in to answer this question.

Browse other questions tagged .