fullseye

vol_frangi — 3D feature op

使い方

3-D Frangi vesselness — multiscale tubular-structure enhancement.

For every sigma in scales the gamma-normalised Hessian is formed, its three eigenvalues |l1| <= |l2| <= |l3| are taken, and the Frangi response

V = (1 - exp(-Ra**2 / 2 alpha**2)) * exp(-Rb**2 / 2 beta**2)
      * (1 - exp(-S**2 / 2 c**2))

is evaluated, where Ra = |l2|/|l3| (plate vs. line), Rb = |l1|/sqrt|l2 l3| (blob deviation) and S = sqrt(l1**2 + l2**2 + l3**2) (structure strength). The response is set to 0 where the contrast polarity is wrong (bright tube: l2 > 0 or l3 > 0; black_ridges=True flips this). The maximum over scales is taken and the volume is normalised to [0, 1].

Parameters

scales : sequence of float — Gaussian sigmas, in voxels, to bracket the vessel radii of interest (see the module “scale-dependent” limitation). alpha, beta : sensitivities of the plate- and blob-suppression terms (Frangi’s defaults 0.5). c : half the maximum Hessian norm S at each scale when None (Frangi’s adaptive suggestion); otherwise a fixed structure-strength scale. black_ridges : False (default) enhances bright tubes on a dark background; True enhances dark tubes on a bright background.

Returns a (D, H, W) float64 volume in [0, 1]. Reference: Frangi et al., MICCAI 1998.

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(feature)

sobel3d · hessian3d · curvature_maps · edt_jfa · vol_sato · vol_hessian_blobness · vol_gradient_magnitude · vol_local_maxima


Provenance: volops.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.