How to combine index arrays and slices in numpy

Question

I have a problem with a combination of index arrays and slices. I have an image (A) and a vector with positions/indexes (pos). Now, I want to select slices (here 3x) from A at different positions. Instead of looping over the positions array, I tried to use the indexer functions (o1), but I does not work. Finally, all slices should be in one array (o2). Can you help me with this problem ?

A = np.array([[0,0,0,0,0,0,3,3],
             [0,0,0,0,0,0,3,3],
             [0,0,0,1,1,0,0,0],
             [0,0,0,1,1,0,0,0],
             [0,0,0,0,0,0,2,2],
             [0,0,0,0,0,0,2,2],
             [0,0,0,0,0,0,0,0],
             [0,0,0,0,0,0,0,0]])

# positions to select
# EDIT:
# pos = np.array([[2,6,7],
#                 [3,4,0]])
pos = np.array([[2,4,0],
                [3,6,6]])

# array with all selections
o1 = np.zeros((3,2,2)).astype(np.int)
# EDIT:
#o1 = A[pos[0]:pos[0]+1,pos[1]:pos[1]+1] ## this gives just one of the values in one area
o1 = A[pos[0]:pos[0] + 2,pos[1]:pos[1] + 2]
print(o1.shape)
print(o1)

# model result
o2 = np.array([[[1,1],[1,1]],[[2,2],[2,2]],[[3,3],[3,3]]])
print(o2.shape)
print(o2)

Another example: With the following line I get the area with the ones. Starting from the position [2,3] choose the next two rows and cols:

print( A[pos[0,0] : pos[0,0] + 2 , pos[1,0] : pos[1,0] + 2]  )

Isn't it possible to extend this to several position pairs (over the whole array pos[]) ?

If np.array([[[1,1],[1,1]],[[2,2],[2,2]],[[3,3],[3,3]]]) is the expected o/p, could you explain how you got that using the inputs? — Divakar
– Divakar, Commented Jun 8, 2017 at 13:36
Your array is in 2 dimensions, why do you use np.zeros with 3 arguments ? — user6547518
– user6547518, Commented Jun 8, 2017 at 13:40
o2 is the expected result. As is mentioned, I have several positions given in pos[] and then I want to use them as index positions in A[] and extract 2x2 parts. In this case [[1,1],[1,1]] , [[2,2],[2,2]] and [[3,3],[3,3]]. All of them should be stored in a singe array (o2). — Jackster182
– Jackster182, Commented Jun 8, 2017 at 13:52

Divakar · Accepted Answer · 2017-06-08 15:38:50Z

1

You could use scikit's sliding window utility to extract such windows given the start indices from pos in a pretty straight forward way -

from skimage.util.shape import view_as_windows

out = view_as_windows(A, (2,2))[pos[0], pos[1]]

Sample run -

In [225]: A
Out[225]: 
array([[0, 0, 0, 0, 0, 0, 3, 3],
       [0, 0, 0, 0, 0, 0, 3, 3],
       [0, 0, 0, 1, 1, 0, 0, 0],
       [0, 0, 0, 1, 1, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 2, 2],
       [0, 0, 0, 0, 0, 0, 2, 2],
       [0, 0, 0, 0, 0, 0, 0, 0],
       [0, 0, 0, 0, 0, 0, 0, 0]])

In [226]: pos = np.array([[2,4,0],[3,6,6]])

In [227]: from skimage.util.shape import view_as_windows

In [228]: view_as_windows(A, (2,2))[pos[0], pos[1]]
Out[228]: 
array([[[1, 1],
        [1, 1]],

       [[2, 2],
        [2, 2]],

       [[3, 3],
        [3, 3]]])

answered Jun 8, 2017 at 15:38

Divakar

222k19 gold badges273 silver badges374 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

John Zwinck · Accepted Answer · 2017-06-08 14:39:42Z

0

It looks like you want this:

A[pos.tolist()]

That is, fancy indexing using pairs of points.

answered Jun 8, 2017 at 14:39

John Zwinck

252k44 gold badges347 silver badges459 bronze badges

2 Comments

Jackster182 Over a year ago

Sry, I do not want just the singe values. I am looking for the whole areas. I changed the line in my first post.

John Zwinck Over a year ago

@Jackster182: Your question is totally unclear. Your use of example input data which is overly large and full of zeros does not help at all. I have no idea what you're asking now.

Collectives™ on Stack Overflow

How to combine index arrays and slices in numpy

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related