pouët.net

new to 3d programming, need help...

category: general [glöplog]
meh, you start learning one thing and all something else falls out of your head :p

Quote:

This is what I had done before I posted here.. the way I'm mapping the vertices is like this :
BB Image
So there are maximum possible 4 vertices for each pixel, one for each corner.

a pic of the mesh (no faces or tris) created from the image set :
BB Image
(you can see there are some odd vertices flying about, I could ignore them but ideally I want to connect them on their own as a separate part of the object, which means the algorithm has to account for that)

added on the 2007-01-20 05:25:29 by chameleon chameleon
Quote:
Hyde: Slices through a torus would have two edges, and you'd simply connect edges that are most similar from slice to slice.

BB Image
It depends on where you slice, so how would you connect the slices above?

However, I think the best proposition in this thread is what you said: use cubes or any other simple primitive. Then relax the blocky mesh to get something "smoother", if the blockyness bothers you.

chameleon: Your point cloud seems very high in density. Why are you wanting to make a mesh out of it. You could use point renderering instead. I forget the exact term, but I think it something like "splatting".
added on the 2007-01-20 09:42:29 by Hyde Hyde
Hyde: Well, yeah there's a special case in the transition between 1 and 2 edges ;). I think it can be dealt with, although, that giant point cloud up there does looks like a sphere.

Still I'd go for the boxes.
added on the 2007-01-20 14:03:13 by doomdoom doomdoom
Quote:
xernobyl: Marching cubes are easier if you have something like a D3D_MarchingCubesPlease() function. Idunno D3D that much. To actually implement marching cubes though is a lot harder than just connecting some dots.


hahaha sorry but you really made me laugh :-).
added on the 2007-01-20 17:13:41 by nystep nystep
polygonizing strange attractors point clouds is really a hard topic, you haven't chosen the easiest in 3d to start with.. good luck though...

random idea: increment a 3d scalar field with each point and use marching cubes over it.
added on the 2007-01-20 17:16:48 by nystep nystep
did you actually try to use marching cubes, or it's big brother marching tetrahedrons? cos seriously, it should give you a decent result especially with a very high res grid, and especially if you manage some distance function from the field rather than binary values.
otherwise look into hughes hoppe and co's work on polygonising point clouds at
http://research.microsoft.com/~hoppe/ .
added on the 2007-01-20 17:59:57 by smash smash

login