layout op• データ種: なし → table(引数だけで決まる op —— 画像やデータの入力を取らない)
• 呼び出し: import tomography; tomography.sinogram_design(n_angles=180, n_detectors=None, size=256, detector_pitch_mm=1.0, span_deg=180.0) (または opstomography.get("sinogram_design"))
What a scan geometry can and cannot resolve — before anything is built.
The axial counterpart of :func:visiondesign.imaging_budget and of
:func:interferometry.csi_design: no data goes in, only the geometry, and
what comes out are the limits that the geometry has already decided.
Returns a dict with, among others:
• `resolvable_feature_mm — 2 * pitch`: two detector samples per cycle is
the Nyquist floor, and no reconstruction algorithm recovers a detail finer
than the detector saw.
• `views_for_full_sampling — ceil(pi/2 * n_detectors)`, the classical
matching of angular to radial sampling (:data:VIEWS_PER_DETECTOR).
• `undersampling_factor — that number over n_angles`. **1.0 or below is
a fully sampled scan.** Above it you are doing sparse-view CT on purpose and
the reconstruction algorithm has to make up the difference; the measured
cost is in this module's docstring and in the test suite's break table.
• `streak_free_radius_px — 1 / d(theta) with d(theta)` in radians:
the radius, in pixels, at which the *azimuthal* sample spacing between
neighbouring views (`r * d(theta)`) grows past one sample. Outside it,
filtered back-projection lays down visible streaks. For 180 views over 180
degrees this is 57.3 px, i.e. a 256-px phantom is already streaking at its
corners. It does not depend on the detector pitch: the radius is
quoted in the same unit the sample spacing is, so the pitch cancels
(`pitch / d(theta)` is the same radius in millimetres, which is what an
earlier version returned under the pixel label — 28.6 "px" at 0.5 mm,
114.6 at 2 mm, for a geometry whose streak radius had not changed).
• `sinogram_bytes / elements` — what you are about to allocate.
• `verdict — "fully sampled" / "sparse view"`.
:param n_angles: planned number of views.
:param n_detectors: planned detector bins; `None` -> enough to cover the
diagonal of a *size* x *size* image.
:param size: reconstruction grid side in pixels.
:param detector_pitch_mm: physical detector bin spacing.
:param span_deg: angular range of the scan.
:returns: dict of floats / ints / str.
:raises ValueError: on non-int counts, non-positive pitch or span.
• 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。
• ct_reconstruction — py -3.11 examples/ct_reconstruction.py
• tomography_reconstruct — py -3.11 examples/tomography_reconstruct.py
table を入力に取れる)—
layout)*Provenance: tomography.py — TOMOGRAPHY 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.