Replacing Numpy elements if condition is met

I have a large numpy array that I need to manipulate so that each element is changed to either a 1 or 0 if a condition is met (will be used as a pixel mask later). There are about 8 million elements in the array and my current method takes too long for the reduction pipeline:

for (y,x), value in numpy.ndenumerate(mask_data): 

    if mask_data[y,x]<3: #Good Pixel
        mask_data[y,x]=1
    elif mask_data[y,x]>3: #Bad Pixel
        mask_data[y,x]=0

Is there a numpy function that would speed this up?

Solution 1:

>>> import numpy as np
>>> a = np.random.randint(0, 5, size=(5, 4))
>>> a
array([[4, 2, 1, 1],
       [3, 0, 1, 2],
       [2, 0, 1, 1],
       [4, 0, 2, 3],
       [0, 0, 0, 2]])
>>> b = a < 3
>>> b
array([[False,  True,  True,  True],
       [False,  True,  True,  True],
       [ True,  True,  True,  True],
       [False,  True,  True, False],
       [ True,  True,  True,  True]], dtype=bool)
>>> 
>>> c = b.astype(int)
>>> c
array([[0, 1, 1, 1],
       [0, 1, 1, 1],
       [1, 1, 1, 1],
       [0, 1, 1, 0],
       [1, 1, 1, 1]])

You can shorten this with:

>>> c = (a < 3).astype(int)

Solution 2:

>>> a = np.random.randint(0, 5, size=(5, 4))
>>> a
array([[0, 3, 3, 2],
       [4, 1, 1, 2],
       [3, 4, 2, 4],
       [2, 4, 3, 0],
       [1, 2, 3, 4]])
>>> 
>>> a[a > 3] = -101
>>> a
array([[   0,    3,    3,    2],
       [-101,    1,    1,    2],
       [   3, -101,    2, -101],
       [   2, -101,    3,    0],
       [   1,    2,    3, -101]])
>>>

See, eg, Indexing with boolean arrays.

Solution 3:

The quickest (and most flexible) way is to use np.where, which chooses between two arrays according to a mask(array of true and false values):

import numpy as np
a = np.random.randint(0, 5, size=(5, 4))
b = np.where(a<3,0,1)
print('a:',a)
print()
print('b:',b)

which will produce:

a: [[1 4 0 1]
 [1 3 2 4]
 [1 0 2 1]
 [3 1 0 0]
 [1 4 0 1]]

b: [[0 1 0 0]
 [0 1 0 1]
 [0 0 0 0]
 [1 0 0 0]
 [0 1 0 0]]

Solution 4:

You can create your mask array in one step like this

mask_data = input_mask_data < 3

This creates a boolean array which can then be used as a pixel mask. Note that we haven't changed the input array (as in your code) but have created a new array to hold the mask data - I would recommend doing it this way.

>>> input_mask_data = np.random.randint(0, 5, (3, 4))
>>> input_mask_data
array([[1, 3, 4, 0],
       [4, 1, 2, 2],
       [1, 2, 3, 0]])
>>> mask_data = input_mask_data < 3
>>> mask_data
array([[ True, False, False,  True],
       [False,  True,  True,  True],
       [ True,  True, False,  True]], dtype=bool)
>>>