layout op• 데이터 종류: 없음 → table(인자만으로 정해지는 연산자 —— 이미지나 데이터 입력을 받지 않습니다)
• 호출: import tomography; tomography.sinogram_design(n_angles=180, n_detectors=None, size=256, detector_pitch_mm=1.0, span_deg=180.0)(또는 opstomography.get("sinogram_design"))
그 스캔 기하로 무엇을 해상할 수 있고 없는가 —— 만들기 전에.
> 아래 상세 설명은 원문입니다 —— 요약과 제목은 번역되어 있습니다.
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.
• 샘플 데이터 카탈로그(DL URL / 라이선스) —— 2-D 는 skimage.data(BSD/public)+ 합성, 3-D 는 실데이터 소스(Stanford/PDS 등)의 DL URL.
• 연산자의 내력·참고문헌 —— 이 연산자 족의 바탕이 된 연구/기법의 출처.
• 알고리즘의 정전(저자·연도)과 용도는 위의 패밀리 사용 가이드에 적혀 있습니다.
• ct_reconstruction — py -3.11 examples/ct_reconstruction.py
• tomography_reconstruct — py -3.11 examples/tomography_reconstruct.py
table 를 입력으로 받는 것)—
layout)*Provenance: tomography.py — TOMOGRAPHY 연산자 레지스트리. 이 op 노트는 tools/opdocs.py md 가 자동 생성합니다(직접 편집하지 마세요).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.