tb_local_std — 2D typed op

数据种类:signalsignal

调用:fullseye.apply(img, "tb_local_std", 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])

tb_local_std: input → output

*图为在 128×128 合成输入上实际运行的输出。左为输入,右为输出。点云以俯视散点显示(亮度 = z),一维序列为折线,体数据为沿 z 的最大值投影,视频为中间帧,复数图像为幅值;无法成像的返回值直接显示数值。*

扫描旋钮 a(0.1 / 0.5 / 0.9,另一旋钮取默认):

tb_local_std: knob a sweep (docs site)

*旋钮 b 不改变输出(实测: 0.1 / 0.5 / 0.9 相同)。*

阶段(前置算子 → 本算子,从左到右):

tb_local_std: stages (docs site)

换别的图像(合成场景 / 照片 / 硬币。上排为输入,下排为对应输出。旋钮取默认):

tb_local_std: other inputs (docs site)

用法

带误差界的滑动标准差。

> 以下的详细说明为原文 —— 摘要与标题已翻译。

The 1-D counterpart of the image operator `local_std` (HALCON's

`deviation_image`): the variance is taken after subtracting the mean

(`E[x^2] - E[x]^2` loses its significant digits on a signal that sits far

from zero), unbiased by `n/(n-1)`, and the residual bias of the square root

removed by `c4(n), so the estimate of sigma` itself is unbiased.

*window* is the number of samples in the sliding window, **rounded up to the

next odd number** so the window can be centred (10 becomes 11). The error

bound below is computed from the window actually used, so the number quoted

stays true. The 2-D side does the same thing — `_k(a)` snaps the knob to

3/5/7/9 — and the typed bridge that exposes this op as `tb_local_std`

scales the knob continuously, so an even value arrives whenever the knob

lands between two odd ones.

**The relative standard error of each estimate is `1/sqrt(2(n-1))`** —

35 % for a 5-sample window, 11 % for 41. Quote it next to any noise figure:

a rolling sigma over 9 samples is +- 25 %, which is wider than most of the

changes people try to read off it.

Returns an array the same length as *x* (the ends are reflected).

Typed bridge of the 1d op `local_std into the 2-D evolution registry: the same implementation, called under the op(v, a, b) convention. a drives window (default 9); b` is unused.

参考(示例数据・文献)

• 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。

在 Studio 中试试

下面的程序已确认可以运行(与图相同的输入)。在 Studio 帮助中,此块会变成按钮,可当场加载并运行。

img_to_signal 0.50 0.50
tb_local_std 0.50 0.50

▸ Load this pipeline  ·  Load & run

可运行的示例(实际调用该算子并已验证的样例)

下面的示例调用的是底层账本算子 local_std。此桥接算子只是把同一实现适配为 fn(v, a, b) 约定,行为完全相同(只是调用形式不同)。

gallery2d_texture_freqpy -3.11 examples/gallery2d_texture_freq.py

类型可衔接的下一个算子(可接受 signal 作为输入)

identity · tb_create_funct_1d_array · tb_smooth_funct_1d_gauss · tb_smooth_funct_1d_mean · tb_derivate_funct_1d · tb_integrate_funct_1d · tb_zero_crossings_funct_1d · tb_abs_funct_1d

同类别(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 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.