提取NumPy数组的索引

时间:2015-04-08 10:43:51

标签: performance numpy scipy scikit-learn scikit-image

我有一个NumPy数组,只包含0和1个元素,如下所示:

import numpy as np
data = np.array([[1, 1 , 0 , 0 , 0 , 0 , 1 , 0],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [0, 0 , 1 , 1 , **1** , 1 , 1 , 0],
                 [0, 0 , 1 , 1 , 1 , 1 , 1 , 1],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [1, 1 , 0 , 0 , 0 , 0 , 0 , 0]])

我必须找出元素1的索引,它在每个方向上被1乘2乘2像素包围。

预期答案的位置以粗体显示。

我正在寻找更简单,更快捷的方法。

2 个答案:

答案 0 :(得分:3)

您可以使用一些signal processing -

import numpy as np
from scipy import signal

# Input
data = np.array([[1, 1 , 0 , 0 , 0 , 0 , 1 , 0],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [0, 0 , 1 , 1 , 1 , 1 , 1 , 0],
                 [0, 0 , 1 , 1 , 1 , 1 , 1 , 1],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [1, 1 , 0 , 0 , 0 , 0 , 0 , 0]])

# Extent of the search           
extent = 2;

# Kernel to be used with 2D convolution                 
kernel = np.ones((2*extent+1,2*extent+1))

# Perform 2D convolution with input data and kernel 
filt_out = signal.convolve2d(data, kernel, boundary='symm', mode='same')

# Find where the convolution resulted in a perfect score, 
# i.e is equal to the number of elements in kernel
R,C = np.where( filt_out == kernel.size )

输出 -

In [66]: print(R,C)
[3] [4]

本节中列出了一种替代方法,ndimage执行与前一种方法相同的卷积,保持其余步骤相同。这是获取卷积输出filt_out -

的代码
import scipy.ndimage
filt_out = scipy.ndimage.convolve(data,kernel)

答案 1 :(得分:2)

使用basic morphological operation

这很容易
import numpy as np
from scipy.ndimage.morphology import binary_erosion


data = np.array([[1, 1 , 0 , 0 , 0 , 0 , 1 , 0],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [0, 0 , 1 , 1 , 1 , 1 , 1 , 0],
                 [0, 0 , 1 , 1 , 1 , 1 , 1 , 1],
                 [1, 1 , 1 , 1 , 1 , 1 , 1 , 0],
                 [1, 1 , 0 , 0 , 0 , 0 , 0 , 0]])

expected = np.array([[0, 0 , 0 , 0 , 0 , 0 , 0 , 0],
                     [0, 0 , 0 , 0 , 0 , 0 , 0 , 0],
                     [0, 0 , 0 , 0 , 0 , 0 , 0 , 0],
                     [0, 0 , 0 , 0 , 1 , 0 , 0 , 0],
                     [0, 0 , 0 , 0 , 0 , 0 , 0 , 0],
                     [0, 0 , 0 , 0 , 0 , 0 , 0 , 0],
                     [0, 0 , 0 , 0 , 0 , 0 , 0 , 0]])

# otherwise known as np.ones((5, 5))
structuring_element = np.array([[1, 1, 1, 1, 1],
                                [1, 1, 1, 1, 1],
                                [1, 1, 1, 1, 1],
                                [1, 1, 1, 1, 1],
                                [1, 1, 1, 1, 1]])

# will be of dtype np.bool but you can convert with .astype(np.int)
# if you really need
result = binary_erosion(data, structuring_element)

print(result)

print(np.allclose(result, expected))