vol_nearest_seed_vector — 3D `medial` op

• Data kinds: voxel → flow_dense

• Call: import fullseye as fs; fs.ledger.vol_nearest_seed_vector(vol_binary, spacing=None) (to call the implementation directly, import volops; volops.vol_nearest_seed_vector(vol_binary, spacing=None); from the registry, ops3d.get("vol_nearest_seed_vector"))

Usage

> This operator's description has not been translated yet. The original text follows as it is.

各 voxel から最近の seed voxel への変位ベクトル `(3, D, H, W)(flow_dense`、成分 dz, dy, dx [voxel])。

seed = 非零 voxel(`> 0.5 で二値化)。seed の上では 0。vol_distance_transform` が返すのは距離の値だけで、

「どの voxel が最近か」は落ちる —— 骨格からの半径(表面 → 骨格の変位)、膜までの肉厚、ESDF の勾配、

ラベルのボロノイ分割はこの向きが要る(2026-09-21、コネクトーム基盤の穴)。`spacing (sz, sy, sx)` を渡すと

最近傍の判定は物理距離で行う(返す変位は voxel 単位のまま —— `|v * spacing|` が物理距離)。

厳密解: Maurer, Qi & Raghavan (IEEE TPAMI 2003) の線形時間 EDT(`scipy.ndimage.distance_transform_edt(return_indices=True)`)。

同距離の seed が複数あるときはその実装の順序で 1 つ(順序に意味を持たせないこと)。seed が 1 つも無いと

ValueError(全 voxel が無限遠 —— 黙って 0 を返さない)。

>>> v = np.zeros((5, 5, 5)); v[2, 2, 0] = 1

>>> d = vol_nearest_seed_vector(v); d.shape, tuple(d[:, 2, 2, 4])

((3, 5, 5, 5), (0.0, 0.0, -4.0))

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)

• nearest_seed_partition — py -3.11 examples_3d/nearest_seed_partition.py

Ops the type connects to (they accept `flow_dense` as input)

fuse_to_voxel

Same category (`medial`)

distance_ridge · skeletonize_vol · medial_axis_points · topology_signature · medial_match · skeleton_junctions3d · skeleton_endpoints3d · skeleton_prune3d

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

vol_nearest_seed_vector — 3D medial op