typed opsignal → imagefullseye.apply(img, "tb_spectrogram", a=0.5, b=0.5) (2-D は 1 画像 + 2 スカラつまみ a,b∈[0,1] のモデル)
図は合成の入力 128×128 で実際に走らせた出力。左が入力、右が出力。点群は上から見た散布(明るさ = z)、1-D 列は折れ線、体積は z 方向の最大値投影、動画は中央フレーム、複素画像は振幅、絵にならない返り値は値そのもの。
つまみ a は出力を変えない(実測: 0.1 / 0.5 / 0.9 で同一)。
つまみ b を振る(0.1 / 0.5 / 0.9、もう一方は既定):

段階(前置きの op → この op。左から順):

別の画像でも(合成シーン / 写真 / 硬貨。上段が入力、下段がその出力。つまみは既定):

STFT magnitude spectrogram -> (freqs, times, S) with S shape
(n_freqs, n_frames). Hann-windowed; hop defaults to win//2.
**Same raw convention as :func:`spectrum`, but a different divisor.** Each
column is the unnormalised ``|rfft(frame * hann(win))|``, so it is not an
amplitude either — and dividing by ``2/win`` is *wrong* here, because the
Hann window has already thrown away part of the signal. The correct one-sided
amplitude conversion divides by the window's coherent gain::
w = np.hanning(win)
amp = S * (2.0 / w.sum()) # bins 1 .. win/2-1; DC / Nyquist: 1/w.sum()
Measured on a unit sine at a bin centre (``rate = 16000`` Hz, 1000 Hz,
amplitude exactly 1.0, ``win = 256``): the raw column peak is
``63.7497786196906``; ``* 2/win`` gives ``0.49804514546633283`` (too small by
exactly the Hann coherent gain ``sum(w)/win = 0.498046875``), while
``* 2/sum(w)`` gives ``0.9999965273676957``. Only the second one is the
amplitude that was actually in the signal.
Peak *positions*, frame-to-frame ratios and any dB *difference* are unaffected
by either factor. This function returns magnitudes only — the phase is
discarded, so it cannot be inverted; use ``acoustics.stft`` / ``acoustics.istft``
for a round-trip.
Typed bridge of the 1d op spectrogram into the 2-D evolution registry: the same implementation, called under the op(v, a, b) convention. a drives rate (default 1) and b drives win (default 256).
下のプログラムは実際に走ることを確かめてある(図と同じ入力)。Studio のヘルプではこのブロックがボタンになり、その場で読み込んで実行できる。
img_to_signal 0.50 0.50
tb_spectrogram 0.50 0.50
次の例は元の台帳 op spectrogram を呼ぶもの。この橋渡し op は同じ実装を fn(v, a, b) 規約に合わせただけなので、挙動はそのまま当てはまる(呼び出し形だけ違う)。
py -3.11 examples/acoustic_condition_monitoring.pyimage を入力に取れる)identity · gaussian · mean_box · bilateral · unsharp · median · min_filter · max_filter
typed)tb_points_to_voxel · tb_estimate_point_normals · tb_iss_keypoints · tb_project_points · tb_render_point_depth · tb_statistical_outlier_removal · tb_radius_outlier_removal · tb_voxel_grid_downsample
Provenance: ops.py — 2D 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.