fullseye

riesz_displacement_series — QUAT motion op

使い方

Whole-frame displacement waveform from the monogenic phase. → (T, 2).

The contrast-weighted spatial mean of :func:riesz_displacement: the rigid-body motion of the scene in pixels, one (dx, dy) row per frame. The (n, 2) layout is the shared 1-D convention (dsp.spectrum, funct1d, opsoptics’s MTF curves), so the trace can be handed to dsp.spectrum to read off a resonant frequency without any repacking.

Weights are the |z|^2 contrast the field solve uses, restricted to the pixels marked valid, so blank regions do not drag the average towards zero. A clip with no valid pixel anywhere (a constant image) returns exact zeros rather than dividing by zero.

Accuracy, the J0 cliff and the multi-orientation failure all inherit from :func:riesz_displacement — read its head-to-head table before trusting a number from here. On a single-grating 64x64x64 / 8 px / 4 Hz clip a true 0.5 px amplitude is recovered as 0.50000000000000 px (2.2e-16 relative); on the two-grating motionmag.synthesize_translation default the same 0.5 px comes back 13.0 % low, with nothing to signal it.

詳しい使い方ガイド

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(pairs を入力に取れる)

同カテゴリ(motion)

riesz_motion_magnify · riesz_displacement


Provenance: quatimage.py — QUAT 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.