mls_smooth — 3D preprocess op

Data kinds: pointspoints

Call: import fullseye as fs; fs.ledger.mls_smooth(points, radius: 'float', order: 'int' = 2) (to call the implementation directly, import pcl_filter; pcl_filter.mls_smooth(points, radius: 'float', order: 'int' = 2); from the registry, ops3d.get("mls_smooth"))

Usage

Drop noise by projecting each point onto a local polynomial surface (Moving Least Squares smoothing).

> The detailed description below is the original text — the summary and the headings are translated.

点ごとに半径 `radius` 内の近傍を集め、重み付き PCA で局所平面(法線 n と接平面の

2 軸)を推定し、近傍の「接平面上座標 (u,v) → 法線方向の高さ h」に order 次の

多項式曲面をガウス重み付き最小二乗で当てはめる。注目点自身は局所座標の原点 (0,0)

にあたるので、当てはめた曲面の (0,0) での高さ(= 定数項)ぶんだけ法線方向へ動かして

曲面上へ射影する。面の形は保ったままセンサノイズだけを均せる。

Parameters

----------

points : array_like, shape (N, 3)

入力点群。

radius : float

近傍球の半径(> 0)。局所曲面のサポート。

order : int

局所多項式の次数(既定 2)。項数は (order+1)(order+2)/2。

Returns

-------

ndarray, shape (N, 3)

平滑後の点群(順序・点数は保持)。近傍が多項式の項数に満たない点は原位置のまま。

Notes

-----

近似手法である。近傍数が項数未満/局所平面が縮退する点は動かさず原位置を維持する

(穴や境界で暴れないための安全策)。`radius <= 0` は ValueError、空入力は空を返す。

Background guides (the physics and conventions behind this op)

depth_sensors — 深度センサの知識 — 測距原理・実機の値・欠測の出方

References (sample data, literature)

• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).

• Operator provenance and references — the sources of the research/methods this op family came from.

Runnable examples (verified samples that actually call this op)

pcl_geodesicpy -3.11 examples_3d/pcl_geodesic.py

Ops the type connects to (they accept points as input)

points_to_voxel · gaussians_to_voxel · estimate_point_normals · to_points · match_points_ncc · match_pca · moment_axes · icp_point2point_3d

Same category (preprocess)

statistical_outlier_removal · radius_outlier_removal · voxel_grid_downsample · volume_downsample


*Provenance: pcl_filter.py — 3D operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*

© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.