如何将一个numpy的二维数组裁剪为非零值? [英] How to crop a numpy 2d array to non-zero values?

问题描述

比方说，我有一个2d布尔numpy数组，如下所示:

Let's say i have a 2d boolean numpy array like this:

import numpy as np
a = np.array([
    [0,0,0,0,0,0],
    [0,1,0,1,0,0],
    [0,1,1,0,0,0],
    [0,0,0,0,0,0],
], dtype=bool)

我通常如何将其裁剪到包含所有True值的最小框(矩形，内核)?

How can i in general crop it to the smallest box (rectangle, kernel) that includes all True values?

因此在上面的示例中:

b = np.array([
    [1,0,1],
    [1,1,0],
], dtype=bool)

推荐答案

经过一番摆弄之后，我本人实际上找到了一个解决方案:

After some more fiddling with this, i actually found a solution myself:

coords = np.argwhere(a)
x_min, y_min = coords.min(axis=0)
x_max, y_max = coords.max(axis=0)
b = cropped = a[x_min:x_max+1, y_min:y_max+1]

以上内容适用于开箱即用的布尔数组.如果您还有其他条件，例如阈值t，并且想要裁剪为大于t的值，只需修改第一行:

The above works for boolean arrays out of the box. In case you have other conditions like a threshold t and want to crop to values larger than t, simply modify the first line:

coords = np.argwhere(a > t)

这篇关于如何将一个numpy的二维数组裁剪为非零值?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！