I just got the motivation to finish an older content tool, and I wanted to share the results here. There are a few scripts in the project, but the main one is a retopology script that is based on Constrained Elastic Surface Nets as presented by Sarah F. F. Gibson (http://www.cs.tufts.edu/~frisken/surfaceNets-Paper.pdf).
This is focused on static mesh retopologizing, not dynamic voxel terrain sculpting. It can be used to assist in generating a game world in a specific way that was useful for us, but is not an in-engine sculpting tool on its own. However, some of the ideas apply, and a lot of topics covered here are important for general in-engine or in-client modeling and creation tools.
One question – it seems there is a rounding error somewhere in initial smoothing step that is causing one of the extreme axis AABB bounds to miss the extreme bounds of points. i.e., if you run CESN against 4-6 tiled planes, one of the edges (the same on each) of each mesh will miss a row/column as it consistently ignores points along the extreme.
I am assuming this because with smoothing off, the edges do line up. However, now that I think about it, it is more likely that these triangles have been excluded from the hitTest and the smoothing step is simply ignoring them completely as it snaps, giving the missing strip effect.
OK – I found the issue. It was in the initial smoothing (point-snapping) pass:
The NearestPointOnTri test is not using all of the triangle vertices, just the first, which is causing the effect to only really show itself on the last row being processed.