fullseye

tree_run_length — GRAPH tree op

• データ種: table × signal → table
• 呼び出し: import fullseye as fs; fs.ledger.tree_run_length(tree, labels, background=0, zero_labels=None) (実装を直接呼ぶなら import treemorph; treemorph.tree_run_length(tree, labels, background=0, zero_labels=None)、台帳から引くなら opsgraph.get("tree_run_length"))

使い方

Expected run length (ERL) of a skeleton under a candidate labelling of its nodes.

labels gives, for every node of the tree table, the candidate object it falls in. A run is a maximal connected piece of the tree whose nodes all carry the same label and that holds at least one edge; its length is the cable inside it (an edge belongs to a run only if both ends share the label; a lone node between two cuts is no run). Edges touching background belong to no run. Runs whose label is in zero_labels (objects that also cover another skeleton — a merge) count as length 0, as in Januszewski et al. 2018: a merged run is not trustworthy anywhere.

erl = sum(l_i^2) / L with L the whole cable — the run length a point drawn uniformly along the skeleton finds itself in (points on cut or background edges find length 0). Also returns run_lengths (one per run, merged runs already zeroed), runs, max_run, cable_length, cut_edges (both ends labelled, differently), background_edges and merged_runs.

Identities: erl == L when every node carries one non-background label; sum(run_lengths) + cut cable + background cable == L; for any labelling with k runs and no lost cable, erl >= L / k (Cauchy-Schwarz) and erl <= max_run; a path of length L cut at m equally spaced points gives exactly L / (m + 1); cut at m uniformly random points, E[erl] = 2L / (m + 2) (Dirichlet(1, .., 1)).

Raises ValueError: not a tree table; labels not one integer per node, non-finite or non-integer.

参考(サンプルデータ・文献)

• サンプルデータ カタログ(DL URL / ライセンス) — 2-D は skimage.data(BSD/public)+ 合成、3-D は実データ源(Stanford/PDS 等)の DL URL。
• 演算子の来歴・参考文献 — この op 族の元になった研究/手法の出典。
• アルゴリズムの正典(著者・年)と用途は上記ファミリ使い方ガイドに記載。

実行できる例(この op を実際に呼ぶ検証済みサンプル)

• poc_skeleton_run_length_vs_voi — py -3.11 examples/poc_skeleton_run_length_vs_voi.py

型が繋がる次の op(table を入力に取れる)

graph_edge_consensus · graph_core_persistence · tree_morphometry · tree_sholl

同カテゴリ(tree)

tree_from_swc · tree_morphometry · tree_sholl

---

Provenance: treemorph.py — GRAPH operator registry. この per-op ノートは tools/opdocs.py md が自動生成(手編集しない)。

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

This site is open source. Improve this page.