fullseye
tb_tcspc_background_subtract — 2D typed op
- データ種:
counts→counts - 呼び出し:
fullseye.apply(img, "tb_tcspc_background_subtract", a=0.5, b=0.5)(2-D は 1 画像 + 2 スカラつまみa,b∈[0,1]のモデル)
図は合成の入力 128×128 で実際に走らせた出力。左が入力、右が出力。点群は上から見た散布(明るさ = z)、1-D 列は折れ線、体積は z 方向の最大値投影、動画は中央フレーム、複素画像は振幅、絵にならない返り値は値そのもの。
つまみ a は出力を変えない(実測: 0.1 / 0.5 / 0.9 で同一)。
つまみ b を振る(0.1 / 0.5 / 0.9、もう一方は既定):
段階(前置きの op → この op。左から順):
別の画像でも(合成シーン / 写真 / 硬貨。上段が入力、下段がその出力。つまみは既定):
使い方
Remove the ambient-light / dark-count floor from an arrival-time histogram.
Outdoors, most of what a dToF sensor counts is sunlight: a roughly uniform
pedestal under the return pulse. It biases the centroid toward the middle of
the window (a floor of ``b`` per bin pulls the first moment toward
``window/2``) and it inflates the apparent signal, so it is removed before
any depth or lifetime estimate.
The level is estimated by *method* and then **subtracted** (the sign trap:
the result is ``hist - level``, clipped at 0, never ``hist + level``):
* ``"median"`` (default) — the median of every bin. Robust while the pulse
occupies well under half the window, which is the normal dToF case.
* ``"leading"`` — the mean of the first *leading_bins* bins, the classical
choice when the pulse is known to arrive late (a far target).
* ``"trailing"`` — the mean of the last *leading_bins* bins, for
fluorescence decays where the tail is background.
* ``"quantile"`` — the given *quantile* of all bins, for tuning by hand.
*leading_bins* defaults to ``None`` = ``min(8, len(hist))``, so the default
call works on a short histogram instead of raising over a constant nobody
chose (a fixed default of 8 made ``method="leading"`` fail on any histogram
with fewer than 8 bins).
*scale* multiplies the estimated level before subtraction (``scale=1.2`` for
a deliberately aggressive removal). Clipping at 0 means the result is a valid
non-negative histogram that the rest of this module will accept.
Ground truth: on a noiseless histogram with a known flat pedestal of 20
counts/bin under a 5000-photon pulse covering 5.1% of the window, the median
estimate recovers 20.000000 and the returned histogram equals the pedestal-
free pulse **exactly** (measured area error 0.0, pinned in the tests).
Returns a float64 1-D histogram of the same length as *hist*.
**Raises** ``ValueError``: negative, non-finite or non-1-D *hist*, an unknown
*method*, a *leading_bins* outside ``[1, len(hist)]``, a *quantile* outside
``[0, 1]``, and a negative *scale*.
Typed bridge of the photon op tcspc_background_subtract into the 2-D evolution registry: the same implementation, called under the op(v, a, b) convention. a drives quantile (default 0.5) and b drives scale (default 1).
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。
Studio で試す
下のプログラムは実際に走ることを確かめてある(図と同じ入力)。Studio のヘルプではこのブロックがボタンになり、その場で読み込んで実行できる。
img_to_counts 0.50 0.50
tb_tcspc_background_subtract 0.50 0.50
実行できる例(この op を実際に呼ぶ検証済みサンプル)
次の例は元の台帳 op tcspc_background_subtract を呼ぶもの。この橋渡し op は同じ実装を fn(v, a, b) 規約に合わせただけなので、挙動はそのまま当てはまる(呼び出し形だけ違う)。
- photon_timeresolved —
py -3.11 examples/photon_timeresolved.py - poc_dtof_ranging —
py -3.11 examples/poc_dtof_ranging.py
型が繋がる次の op(counts を入力に取れる)
identity · tb_spad_deadtime_apply · tb_spad_deadtime_correct · tb_tcspc_coates_correct · tb_tcspc_irf_convolve · tb_dtof_depth · tb_countrate_to_counts · tb_counts_to_countrate
同カテゴリ(typed)
tb_points_to_voxel · tb_estimate_point_normals · tb_iss_keypoints · tb_project_points · tb_render_point_depth · tb_statistical_outlier_removal · tb_radius_outlier_removal · tb_voxel_grid_downsample
Provenance: ops.py — 2D 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.