Extract random 2d windows of a 2d numpy array

问题

import numpy as np
arr = np.array(range(60)).reshape(6,10)

arr

> array([[ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9],
>        [10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
>        [20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
>        [30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
>        [40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
>        [50, 51, 52, 53, 54, 55, 56, 57, 58, 59]])

What I need:

select_random_windows(arr, number_of windows= 3, window_size=3)



>  array([[[ 1,  2,  3],
>            [11, 12, 13],
>            [21, 22, 23]],
>            
>            [37, 38, 39],
>            [47, 48, 49],
>            [57, 58, 59]],
>           
>            [31, 32, 33],
>            [41, 42, 43],
>            [51, 52, 53]]])

In this hypothetical case I'm selecting 3 windows of 3x3 within the main array (arr).

My actual array is a raster and I basically need a bunch (on the thousands) of little 3x3 windows.

Any help or even a hint will be much appreciated.

I actually haven't found any practical solution yet...since many many hours

THX!

回答1:

We can leverage np.lib.stride_tricks.as_strided based scikit-image's view_as_windows to get sliding windows. More info on use of as_strided based view_as_windows.

from skimage.util.shape import view_as_windows

def select_random_windows(arr, number_of_windows, window_size):
    # Get sliding windows
    w = view_as_windows(arr,window_size)

    # Store shape info
    m,n =  w.shape[:2]

    # Get random row, col indices for indexing into windows array
    lidx = np.random.choice(m*n,number_of_windows,replace=False)
    r,c = np.unravel_index(lidx,(m,n))
    # If duplicate windows are allowed, use replace=True or np.random.randint

    # Finally index into windows and return output
    return w[r,c]

Sample run -

In [209]: arr
Out[209]: 
array([[ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
       [20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
       [30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
       [40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
       [50, 51, 52, 53, 54, 55, 56, 57, 58, 59]])

In [210]: np.random.seed(0)

In [211]: select_random_windows(arr, number_of_windows=3, window_size=(2,4))
Out[211]: 
array([[[41, 42, 43, 44],
        [51, 52, 53, 54]],

       [[26, 27, 28, 29],
        [36, 37, 38, 39]],

       [[22, 23, 24, 25],
        [32, 33, 34, 35]]])

回答2:

You can try [numpy.random.choice()][1]. It takes a 1D or an ndarray and creates a single element or an ndarray by sampling the elements from the given ndarray. You also have an option of providing the size of the array you want as the output.

来源：https://stackoverflow.com/questions/52642035/extract-random-2d-windows-of-a-2d-numpy-array