http://faculty.cs.tamu.edu/schaefer/res ... zation.pdf

What do you guys think about this method to extract the surface as an alternative to Marching cubes? I've glanced at their running time and thought I could definitely live with those extract times. Here I assume coeff. means time taken to generate the wavelet coefficients and synthesis means extraction.

Edit: I didn't read the paper closely, but when they say quadtree do they mean a implicit quadtree process? Bottom up approach or a top down? I've been drinking a little so hopefully I'm making some sense.

I only skimmed it but I think you misunderstood. As far as I can tell they are only describing an algorithm for converting a mesh into voxels, rather than the other way around. The reference to Marching Cubes is simply that they are using it to judge the quality of thier voxelisation. I think the times for both coefficients and synthesis relate to different parts of the voxelisation process.

By the way, there's a page on the Wiki to collect interesting links to papers such as this. It's a bit empty (just added yours) but feel free to add any other interesting papers you come across. You can also post in the forum to discuss them of course.

Oh I see now. Rasterization of triangular mesh. I guess it's still tangentially related to PolyVox if you need to get a mesh into the volume, that is if you got the time to implement their algorithm.

Oh it's defintely related and useful. The castle in the video on the Thermite3D front page was made by a kind of rasterisation of meshes into a volume. My algorithm was pretty poor compared to theirs (it just kept subdividing a triangle until it was smaller that a voxel, then wrote to that voxel) but it got the job done