fullseye

patch_features — SPC anomaly_patch op

• データ種: image2d → table
• 呼び出し: import fullseye as fs; fs.ledger.patch_features(image, patch=8, stride=4, scales=(1.0, 2.0, 4.0), groups=('gradient', 'gabor', 'lbp'), n_orient=4) (実装を直接呼ぶなら import anomaly_patch; anomaly_patch.patch_features(image, patch=8, stride=4, scales=(1.0, 2.0, 4.0), groups=('gradient', 'gabor', 'lbp'), n_orient=4)、台帳から引くなら opsspc.get("patch_features"))

使い方

Classical patch features on a grid: one D-vector per patch (table).

image is (H, W) or (H, W, 3). Pixel channels are computed at every Gaussian scale in scales (sigma, px) and averaged over patch x patch windows every stride px: intensity (smoothed mean), gradient (Gaussian-derivative magnitude and the doubled-angle components |g|cos2t, |g|sin2t: the structure tensor), gabor (zero-mean Gabor energy, wavelength 4 sigma, n_orient orientations), lbp (histogram of the 10 rotation-invariant uniform local binary patterns, radius = round(sigma)), color (red-green and yellow-blue opponents, needs a colour image). fourier is computed on the raw patches, not per scale: the log share of a Hann-windowed patch’s spectrum in three radial bands. Returns features (gh, gw, D), names (D), grid (gh, gw), dim and the parameters.

Raises ValueError on an empty / non-finite / wrongly shaped image, an image smaller than patch, unknown groups, or color on a gray image.

参考(サンプルデータ・文献)

• サンプルデータ カタログ(DL URL / ライセンス) — 2-D は skimage.data(BSD/public)+ 合成、3-D は実データ源(Stanford/PDS 等)の DL URL。
• 演算子の来歴・参考文献 — この op 族の元になった研究/手法の出典。
• アルゴリズムの正典(著者・年)と用途は上記ファミリ使い方ガイドに記載。

実行できる例(この op を実際に呼ぶ検証済みサンプル)

• poc_patch_anomaly — py -3.11 examples/poc_patch_anomaly.py

型が繋がる次の op(table を入力に取れる)

msa_anova_table · msa_gauge_rr · msa_bias_linearity · msa_attribute_agreement · gum_standard_uncertainty · gum_propagate · gum_expanded · gum_monte_carlo

同カテゴリ(anomaly_patch)

patchdist_fit · patchdist_score · patchbank_fit · patchbank_score

---

Provenance: anomaly_patch.py — SPC 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.

This site is open source. Improve this page.