counting op• 資料種類:image2d → table
• 呼叫:import photoncount; photoncount.photon_statistics(counts)(或 opsphoton.get("photon_statistics"))
光子計數影格的卜瓦松統計:它真的是散粒雜訊受限的嗎?
> 以下的詳細說明為原文 —— 摘要與標題已翻譯。
Returns a dict: `mean · variance (population, ddof=0`) ·
`fano_factor = variance / mean` (1 for a Poisson process) ·
`snr_poisson = sqrt(mean)` (the theoretical photon-limited SNR) ·
`snr_measured = mean / std` (what this frame actually achieved) ·
`total_counts · n_samples · zero_fraction` (the fraction of pixels
that saw no photon at all — the honest measure of "photon starved";
`exp(-lambda) for a flat field) · max_counts`.
The Fano factor is evidence of Poisson statistics only on a flat field.
On a structured scene the scene's own spatial variance dominates and the
ratio is large and meaningless — this op computes the number, it cannot tell
you which situation you are in. Measured on the test scenes: a flat
`lambda = 100` field (512x512, seed 0) gives 1.001089; the same detector
looking at a linear ramp from 20 to 180 photons gives 22.4102. Both are
"correct" and only one of them means anything.
Raises `ValueError`: negative, non-finite or non-2-D *counts*, fewer
than 2 pixels (no variance), an all-zero frame (`fano_factor` would be
`0/0` — say "no photons were detected" instead of returning NaN), and a
frame with exactly zero variance (`snr_measured would be inf`; for
`n >= 2` a constant frame is not a Poisson realisation but a synthetic
constant, i.e. an input mistake).
• 範例資料目錄(下載 URL / 授權) —— 2-D 用 skimage.data(BSD/公有領域)加合成圖,3-D 給出真實資料源(Stanford/PDS 等)的下載 URL。
• 運算子來歷與參考文獻 —— 該運算子族所依據的研究/方法出處。
• 演算法的正典(作者・年份)與用途見上面的族使用指南。
• photon_timeresolved — py -3.11 examples/photon_timeresolved.py
table 作為輸入)—
counting)photon_sample · photon_uncertainty
*Provenance: photoncount.py — PHOTON 運算子登記表。本條目由 tools/opdocs.py md 自動產生(請勿手動編輯)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.