How can I Slice a 2D NumPy Array into Smaller 2D Arrays?

Slicing 2D Arrays into Smaller 2D Arrays in NumPy

When working with images or other data represented as 2D arrays, it is often necessary to slice them into smaller arrays for processing or analysis. In numpy, this can be achieved using a combination of reshape and swapaxes operations.

To understand the approach, let's consider an example. Suppose we have a 2x4 array c:

c = np.arange(24).reshape((4, 6))

and we want to slice it into two 2x2 arrays. We know that the first block should comprise the elements from the first two rows and first two columns, and the second block from the remaining elements.

The reshape operation can be used to reshape the array into a new shape, in this case 2x2x3. The number of rows and columns in the new shape is specified as the first two arguments, and the third argument, -1, tells numpy to calculate the necessary dimension to make the reshape valid.

c.reshape(2, 2, -1)

swapaxes then swaps the rows and columns to get the desired block format:

c.reshape(2, 2, -1).swapaxes(1, 2)

Finally, a second reshape is used to flatten the array into the desired shape:

c.reshape(2, 2, -1).swapaxes(1, 2).reshape(-1, 2, 2)

The result is two 2x2 arrays, as desired:

[[[ 0  1]
  [ 6  7]] [[ 2  3]
  [ 8  9]]]

[[[12 13]
  [18 19]] [[14 15]
  [20 21]]]

The provided code includes a function, blockshaped, that generalizes this slicing approach for any desired number of blocks:

def blockshaped(arr, nrows, ncols):
    return (arr.reshape(h//nrows, nrows, -1, ncols)
               .swapaxes(1, 2)
               .reshape(-1, nrows, ncols))

The above is the detailed content of How can I Slice a 2D NumPy Array into Smaller 2D Arrays?. For more information, please follow other related articles on the PHP Chinese website!