gray opvoxel → voxelimport fullseye as fs; fs.ledger.vol_equalize(vol, nbins=256, mask=None) (実装を直接呼ぶなら import volgray; volgray.vol_equalize(vol, nbins=256, mask=None)、台帳から引くなら ops3d.get("vol_equalize"))Histogram equalisation of a volume (HALCON equ_histo_image).
Builds an nbins-bin histogram over the volume’s [min, max] range,
takes its cumulative distribution as a monotone LUT, and maps every voxel
through it — the output lives in (0, 1] and its histogram is
(approximately) flat. With mask (thresholded at > 0.5, the volops
convention) the histogram is computed from the masked voxels only while
the LUT is applied to the whole volume — exactly the HALCON
reduce_domain + equ_histo_image domain behaviour, and the right
tool when a dominant background (e.g. air around a CT subject) would
otherwise swallow the whole dynamic range.
A constant volume is returned unchanged — normalising a flat volume would amplify floating-point dust into full-scale garbage (fail-honest, not fail-loud; see the module docstring).
vol : array_like, shape (D, H, W)
Input volume (coerced to float64; NaN/Inf rejected).
nbins : int
Histogram bins, >= 2. More bins = finer LUT (the mapping is
piecewise-constant per bin — an approximation, documented above).
mask : array_like, shape (D, H, W), optional
Histogram domain. Must match vol’s shape and select at least one
voxel (an empty domain has no histogram — fail-closed).
ndarray, shape (D, H, W), float64
The equalised volume in [0, 1] (constant input: the input itself).
ValueError
Non-3-D / non-finite input, nbins < 2, a mask shape mismatch, or
an empty mask.
py -3.11 examples_3d/gray_window_level.pyvoxel を入力に取れる)voxel_to_mips · voxel_to_mesh · signed_distance_field · to_points · sobel3d · hessian3d · curvature_maps · edt_jfa
gray)vol_window_level · vol_gamma · vol_stretch
Provenance: volgray.py — 3D 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.