fidelity op• Data kinds: image2d × image2d → scalar
• Call: import fullseye as fs; fs.ledger.mse(a, b) (to call the implementation directly, import imgmetrics; imgmetrics.mse(a, b); from the registry, opsimgmetrics.get("mse"))
Mean squared error. A raw quantity that does not depend on `data_range`.
> The detailed description below is the original text — the summary and the headings are translated.
式: `mean((a - b)^2)`(全要素、float64 で計算)。
• `a, b: 同じ形なら次元は問わない(2-D グレー、(H, W, 3)`、3-D 体積)。
dtype も問わないが 画素値の尺度をそのまま引き継ぐ ―― uint8 の 2 枚と、
それを `/255 した float の 2 枚では値が 255^2` 倍違う。尺度を揃えて比べたい
なら `psnr(data_range` で正規化)を使う。
• 返り値: Python の `float。0 が完全一致、上限は data_range^2`。
• 失敗(`MetricContractError`): 形が違う / 空 / NaN・Inf を含む(黙って伝播させない)。
平方根が要るなら `rmse。まとめて測るなら compare_images`。
• image_difference_metrics family guide
• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).
• Operator provenance and references — the sources of the research/methods this op family came from.
• The canonical algorithm (author, year) and its uses are named in the family usage guide above.
• image_quality_metrics — py -3.11 examples/image_quality_metrics.py
scalar as input)—
fidelity)rmse · psnr · ssim · ms_ssim · ssim_map
*Provenance: imgmetrics.py — IMGMETRICS operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.