Mesh scale factor

[tibuch]

Mesh scale factor

[odinsbane]

Hello

I have wondered about using the marching cubes algorithm before. I didn't know if it would produce produce simple closed surfaces. It sounds like a good idea though!

The scaling always uses the long axis of the image.

Lx = labeling.shape[2]*spacing[2]
Ly = labeling.shape[1]*spacing[1]
Lz = labelling.shape[0]*spacing[0]

(I suspect that is how python is ordering the axis, but if not I hope it makes sense.)

Then the scale factor will be

factor = 1/max(Lx, Ly, Lz)

The offset is half the image as you have noticed.

ox = -0.5*Lx*factor
oy = -0.5*Ly*factor
oz = -0.5*Lz*factor

So you need to know the spacing and the image dimensions. I hope that helps.

Cheers

[tibuch]

Awesome thanks!

I think the step_size parameter in marching cubes will be crucial to generate simple meshes. With step_size = 1 the meshes are very dense and deforming them takes longs. However, if I use connection remesh the meshes get simple enough. I will experiment now with increasing step_size.

That being said, using these meshes as initialization works very well!

[tibuch]

Just in case, I put my script to convert a label image to meshes here: https://github.com/fmi-faim/faim-3d-organoid-seg/tree/main/ipa/convert-labeling-to-meshes

Using the step_size parameter in marching cubes works well to get coarse meshes. One just has to be careful to handle objects which are smaller than the step_size.