ct_scan op• 数据种类:projstack × ctgeometry → voxel
• 调用: import fullseye as fs; fs.ledger.tv_reconstruct(projections, geometry, tv_weight=0.0003, n_iter=200, initial=None, nonnegative=True, device='auto', on_iter=None)(要直接调用实现,import ctscan; ctscan.tv_reconstruct(projections, geometry, tv_weight=0.0003, n_iter=200, initial=None, nonnegative=True, device='auto', on_iter=None);从台账取用则 opstomography.get("tv_reconstruct"))
• GPU:无(仅 CPU 参考实现)
基于 Chambolle-Pock 原始对偶法的全变分重建。
> 以下的详细说明为原文 —— 摘要与标题已翻译。
Solves min 0.5 ||A x - p||^2 + w ||grad x||_1 with x >= 0, in the form Sidky, Jorgensen & Pan 2012 give for CT. The prior says that the object is piecewise constant, which is what lets it fill data that was never measured: few views and limited angle. On an object that is not piecewise constant it paints the same flat patches, so a low residual here is not evidence of truth.
*tv_weight* is in units of ||A||^2 times the volume scale, so that one value behaves alike across grid sizes.
:param projections: (view, V, U) line integrals. :param geometry: dict from ct_scan_geometry. :param tv_weight: strength of the prior, >= 0. :param n_iter: primal-dual iterations, 1 .. 5000. :param initial: starting volume; None gives zeros. :param nonnegative: enforce x >= 0. :param device: "auto", "cpu" or "cuda". :param on_iter: optional on_iter(iteration, volume). :returns: (Z, Y, X) float volume. :raises ValueError: on a shape that disagrees with the geometry or a negative
weight. Empty input (projections with 0 elements) raises ValueError.
• 示例数据目录(下载 URL / 许可证) —— 2-D 用 skimage.data(BSD/公有领域)加合成图,3-D 给出真实数据源(Stanford/PDS 等)的下载 URL。
• 算子来历与参考文献 —— 该算子族所依据的研究/方法出处。
• 算法的正典(作者、年份)与用途见上面的族使用指南。
• ct_scan_modes — py -3.11 examples/ct_scan_modes.py
voxel 作为输入)radon_volume · ct_forward_project
ct_scan)ct_scan_geometry · shepp_logan_3d · ellipsoid_phantom · ellipsoid_projections · ct_forward_project · ct_adjoint_project · fdk_reconstruct · sirt_reconstruct
*Provenance: ctscan.py — TOMOGRAPHY 算子登记表。本条目由 tools/opdocs.py md 自动生成(请勿手工编辑)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.