Numpy find average of every 2x2 array inside a large array

Question

I have a 6x6 numpy array

From this I want to find the average of every 2x2 array inside the large 6x6 array. Like:

array= 
[10, 20, 30, 20, 40, 15
 34, 56, 67, 34, 45, 91
 23, 87, 56, 90, 34, 44
 10, 20, 30, 12, 15, 15
 34, 56, 67, 11, 10, 91
 23, 87, 56, 90, 34, 44]


set1= 
[10, 20
34, 56]  Average=30

set2=
[20, 30
56, 67]  Average=43

....
....

setN=
[10, 91
34, 44]  Average=44

Thanks in advance.

Welcome to StackOverflow. Regardless of the answers already given, you need to post your attempt at solving the problem and not expect people to write code for you. — HazardousGlitch, Jul 20 '20 at 16:34

Jan Christoph Terasa · Answer 1 · 2020-07-20T17:08:57.597

Essentially you want to calculate the 2D moving average of your array, with a 2x2 window. This can be done by convolution, scipy.signal.convolve2d can calculate a 2D convolution.

First, you should define your array to actually be 2D:

import numpy as np

arr = np.array(
[[10, 20, 30, 20, 40, 15],
 [34, 56, 67, 34, 45, 91],
 [23, 87, 56, 90, 34, 44],
 [10, 20, 30, 12, 15, 15],
 [34, 56, 67, 11, 10, 91],
 [23, 87, 56, 90, 34, 44]])

Then do a 2D convolution using scipy.signal.convolve2d:

from scipy.signal import convolve2d

# create an averaging kernel
kernel = np.ones((2, 2), dtype=float)
kernel /= kernel.size

# compute 2d moving average
conv = convolve2d(arr, kernel, mode='valid')
print(conv[0][0]) # 30.

If you want to get the result as dtype=int, do:

conv = conv.astype(int)

You can also check this answer.

Sorry, please ask your questions here. If this is too advanced, you should find a simpler solution you can understand yourself. — Jan Christoph Terasa, Jul 20 '20 at 16:23
I added some more links to related topics to help you further. — Jan Christoph Terasa, Jul 20 '20 at 16:27

Nicolas Gervais · Answer 2 · 2020-07-20T16:23:40.993

You can use skimage.util.shape.view_as_windows:

import numpy as np
from skimage.util.shape import view_as_windows
array = np.array([[10, 20, 30, 20, 40, 15],
                 [34, 56, 67, 34, 45, 91],
                 [23, 87, 56, 90, 34, 44],
                 [10, 20, 30, 12, 15, 15],
                 [34, 56, 67, 11, 10, 91],
                 [23, 87, 56, 90, 34, 44]])

My implementiation:

b = view_as_windows(array, window_shape=(2, 2), step=1)

np.array(list(map(lambda x: [np.mean(i) for i in x], b)))

Result:

array([[30.  , 43.25, 37.75, 34.75, 47.75],
       [50.  , 66.5 , 61.75, 50.75, 53.5 ],
       [35.  , 48.25, 47.  , 37.75, 27.  ],
       [30.  , 43.25, 30.  , 12.  , 32.75],
       [50.  , 66.5 , 56.  , 36.25, 44.75]])

Your step size is too large, you should only move by one step, not by two. — Jan Christoph Terasa, Jul 20 '20 at 16:21

Michael Green · Answer 3 · 2020-07-20T16:07:46.630

0

You could do something like:

m, n = 2, 2

avgs = [
    array[x:x+m,y:y+n].mean()
    for x in range(array.shape[0]-m) 
    for y in range(array.shape[1]-n)
]

This will raster across your array collecting the average value in the MxN subarrays

edited Jul 20 '20 at 16:07

answered Jul 20 '20 at 16:01

Michael Green

719
6
15

1

`min(avgs)` but you really should show an attempt at solving this yourself when you ask a question, otherwise it's [off-topic](https://stackoverflow.com/help/on-topic). – ggorlen Jul 20 '20 at 16:09
@AdarshBabu You can use the logic here to create a function that stores (m, n) if it's the current lowest value, then return (m, n) once done – Michael Green Jul 20 '20 at 16:12

Numpy find average of every 2x2 array inside a large array

3 Answers3