Indexing numpy arrays with numpy arrays, in arbitrary dimension

Question

Suppose img is a 2 dimensional numpy array. Suppose also that x and y are integer valued 2 dimensional numpy arrays of the same shape as img. Consider:

newImg = img[x, y]

newImg is now a 2 dimensional array of the same shape as img where newImg[i,j] == img[ x[i,j], y[i,j] ] for all i and j.

I want to generalize this procedure to an arbitrary number of dimensions. That is, let img be a d-dimensional numpy array and take x[i], for i in range(0, d), to be an integer valued d-dimensional numpy array of the same shape as img. What I basically want is:

newImg = img[x[0], x[1], ..., x[d-1]]

Which is obviously pseudocode, and not expected to work.

How can I do this with NumPy?

wim · Accepted Answer · 2014-11-04 00:14:36Z

1

Have you tried simply

newImg = img[x]

It looks like this should work!

I'm assuming x is a list or tuple of integer arrays satisfying the following conditions

len(x) == img.ndim
all(a.shape == img.shape for a in x)

which seems to match what you are describing.

edited Nov 4, 2014 at 0:14

answered Nov 4, 2014 at 0:09

wim

368k114 gold badges681 silver badges816 bronze badges

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

3 Comments

NLi10Me Over a year ago

I don't think it's that simple. There's an example below (it was too big to fit in this comment.)

NLi10Me Over a year ago

Oh, wait, I haven't tried x as a tuple of arrays. I'll try that out.

NLi10Me Over a year ago

Sorry, the tuple of arrays works. Note, that is the solution. In my question, I was assuming x was an array of arrays. The tuple works.

Collectives™ on Stack Overflow

Indexing numpy arrays with numpy arrays, in arbitrary dimension

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related