variation op• Datenarten: image2d × table → table
• Aufruf: import fullseye as fs; fs.ledger.varmodel_score(image, model, mask=None, align='check') (die Implementierung direkt: import varmodel; varmodel.varmodel_score(image, model, mask=None, align='check'); aus dem Register: opsspc.get("varmodel_score"))
Ein Bild gegen ein varmodel_fit-Modell bewerten: z-Karte in Sigma, Karte außerhalb des Bands und eine Erklärung je Region (table).
> Die ausführliche Beschreibung unten ist der Originaltext — Zusammenfassung und Überschriften sind übersetzt.
z = (image - center) / spread on valid pixels (0 elsewhere); a pixel is out of band when |z| > k. Out-of-band pixels are grouped into 8-connected regions; each region reports area, peak_z (signed, the largest |z|), mean_abs_z, polarity (bright / dark / mixed), centroid (row, col), bbox (r0, c0, r1, c1, exclusive end) and peak_pos; regions are sorted by |peak_z| (rounded to 1e-9, ties by label order). Also returns z, exceed, valid (scored pixels, image grid), n_exceed, n_valid, k, far, expected_false = far * n_valid and count_pvalue = P(Binomial(n_valid, far) >= n_exceed) — that p-value assumes independent pixels and is optimistic for spatially correlated noise (reported, not used to decide).
Alignment: the image's translation against the model centre is measured as in varmodel_alignment and returned as offset. align="check" REFUSES an image displaced by more than the model's max_misalign_px; align="translate" registers it to the model by a Fourier shift, computes z in model coordinates and maps z / exceed back onto the image grid by the rounded offset (the wrapped border is invalid), so the maps line up with masks drawn on the original image. mask is on the image grid.
Raises ValueError when image is empty / not 2-D / non-finite, its shape differs from the model, model lacks the fitted keys, mask has the wrong shape, align is unknown, or the image is misaligned (check) / displaced by more than a quarter of the image (translate).
• Katalog der Beispieldaten (Download-URLs / Lizenzen) — 2-D nutzt skimage.data (BSD/Public Domain) plus synthetische Bilder, 3-D nennt Download-URLs echter Datenquellen (Stanford, PDS, …).
• Herkunft und Literatur der Operatoren — die Quellen der Forschung/Verfahren, auf denen diese Operatorfamilie beruht.
• Der kanonische Algorithmus (Autor, Jahr) und seine Anwendungen stehen im Familienleitfaden oben.
• poc_good_sample_variation_model — py -3.11 examples/poc_good_sample_variation_model.py
• poc_patch_anomaly — py -3.11 examples/poc_patch_anomaly.py
table als Eingabe)msa_anova_table · msa_gauge_rr · msa_bias_linearity · msa_attribute_agreement · gum_standard_uncertainty · gum_propagate · gum_expanded · gum_monte_carlo
variation)varmodel_alignment · varmodel_band_k · varmodel_fit · varmodel_calibrate
*Provenance: varmodel.py — SPC Operator-Registry. Diese Notiz wird von tools/opdocs.py md erzeugt (nicht von Hand bearbeiten).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.