PoissonMesh

class PoissonMesh(mesh, repair=True)

Bases: TriangularMesh

Attributes

Methods

PoissonMesh.fit(positions[, voxel_size, ...])

Poisson surface reconstruction.

classmethod fit(positions, voxel_size=1, depth=9, k_neighbors=50, deldist=1.5, density_quantile=0.0, n_threads=1, normals=None, **kwargs)

Poisson surface reconstruction.

Parameters:

positionsnp.ndarray

Point coordinates with shape (n, 3).

voxel_sizefloat

Sampling rate of the input point cloud.

depthint

Depth of the octree used for reconstruction. Higher values capture finer detail but are slower.

k_neighborsint

Number of neighbors for normal estimation. Ignored when normals is given.

deldistfloat

Drop mesh vertices further than this distance from the input point cloud (in original coordinates).

density_quantilefloat

Remove vertices with density below this quantile (0-1). Helps clean low-confidence regions. Set to 0 to disable.

n_threadsint

Number of threads for Poisson reconstruction. Set to -1 to use all available cores.

normalsnp.ndarray, optional

Precomputed per-point normals with shape (n, 3). When provided, internal normal estimation is skipped.

Returns:

PoissonMesh

Reconstructed surface mesh.