typed op• 数据种类:video → video
• 调用:fullseye.apply(img, "tb_temporal_bandpass", a=0.5, b=0.5)(2-D 的模型是一张图 + 两个标量旋钮 a,b∈[0,1])
对每个像素的时间序列进行理想时域带通滤波 -> `(T, H, W)`。
> 以下的详细说明为原文 —— 摘要与标题已翻译。
Each pixel is transformed along time, every DFT bin whose frequency lies
outside `[f_lo, f_hi]` (in Hz, magnitude, DC always excluded) is zeroed,
and the result is transformed back. Frequency-selective where
`videops.moving_average and videops.spatiotemporal_gaussian` are
low-pass; this is the filter isolating "what is happening at 4 Hz".
Exact for a component sitting on a bin: with `T frames at fps`, a
sinusoid at `k*fps/T` Hz passes with gain 1 and everything else in the band
passes untouched. Measured on a bin-centred 4 Hz unit sinusoid in a 64-frame
32 fps clip that also carries a DC offset of 0.5 and a 12 Hz component of
amplitude 0.3, the recovered waveform matches the 4 Hz term alone to
`max|err| = 4.36e-15`.
A brick-wall filter rings in time; that is the price of an exact pass-band
and it is the same choice the 2012 Eulerian magnification paper makes. The
output is zero-mean along time by construction.
Typed bridge of the motionmag op `temporal_bandpass into the 2-D evolution registry: the same implementation, called under the op(v, a, b) convention. This op has no tunable parameter; a and b` are unused.
• 示例数据目录(下载 URL / 许可证) —— 2-D 用 skimage.data(BSD/公有领域)加合成图,3-D 给出真实数据源(Stanford/PDS 等)的下载 URL。
• 算子来历与参考文献 —— 该算子族所依据的研究/方法出处。
• (暂无)
video 作为输入)identity · tb_temporal_band_power · tb_temporal_median_window · tb_moving_average_window · tb_background_subtraction_window · tb_frame_difference_causal · tb_exponential_background · tb_exponential_foreground
typed)tb_points_to_voxel · tb_estimate_point_normals · tb_iss_keypoints · tb_angle_3points · tb_project_points · tb_render_point_depth · tb_statistical_outlier_removal · tb_radius_outlier_removal
*Provenance: ops.py — 2D 算子登记表。本条目由 tools/opdocs.py md 自动生成(请勿手工编辑)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.