fullseye

photon_uncertainty — PHOTON counting op

使い方

Per-pixel Poisson error bar of a photon-count frame.

For a Poisson variable the variance equals the mean, so the one-sigma uncertainty of a single measurement N is sqrt(N) — no calibration, no noise model to fit. With relative=True the returned map is the relative uncertainty 1/sqrt(N) instead (its reciprocal is the per-pixel SNR).

zero_floor replaces counts below it before the square root. It exists because N = 0 gives sqrt(0) = 0, i.e. “this pixel is exactly zero with no uncertainty”, which is wrong: the 95% Poisson upper limit for a single observed zero is about 3 photons. Set zero_floor=1.0 for the common “one-count prior” convention. It is not applied silently — the default is 0.0 and the absolute map really does return 0 there.

Returns a float64 (H, W) image.

Raises ValueError: negative, non-finite or non-2-D counts, a negative zero_floor, and — instead of returning infrelative=True with any pixel at 0 after the floor (that is the division 1/sqrt(0); pass zero_floor > 0 to say what a zero should mean).

詳しい使い方ガイド

参考(サンプルデータ・文献)

実行できる例(この op を実際に呼ぶ検証済みサンプル)

型が繋がる次の op(image2d を入力に取れる)

photon_sample · photon_statistics · anscombe_transform · anscombe_inverse

同カテゴリ(counting)

photon_sample · photon_statistics


Provenance: photoncount.py — PHOTON 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.