fullseye

vol_equalize — 3D gray op

使い方

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).

Parameters

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).

Returns

ndarray, shape (D, H, W), float64 The equalised volume in [0, 1] (constant input: the input itself).

Raises

ValueError Non-3-D / non-finite input, nbins < 2, a mask shape mismatch, or an empty mask.

参考(サンプルデータ・文献)

実行できる例(この op を実際に呼ぶ検証済みサンプル)

型が繋がる次の op(voxel を入力に取れる)

voxel_to_mips · voxel_to_mesh · signed_distance_field · to_points · sobel3d · hessian3d · curvature_maps · edt_jfa

同カテゴリ(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.