import numpy as np
def blockshaped(arr, nrows, ncols):
"""
Return an array of shape (n, nrows, ncols) where
n * nrows * ncols = arr.size
If arr is a 2D array, the returned array looks like n subblocks with
each subblock preserving the "physical" layout of arr.
"""
h, w = arr.shape
return (arr.reshape(h//nrows, nrows, -1, ncols)
.swapaxes(1,2)
.reshape(-1, nrows, ncols))
def unblockshaped(arr, h, w):
"""
Return an array of shape (h, w) where
h * w = arr.size
If arr is of shape (n, nrows, ncols), n sublocks of shape (nrows, ncols),
then the returned array preserves the "physical" layout of the sublocks.
"""
n, nrows, ncols = arr.shape
return (arr.reshape(h//nrows, -1, nrows, ncols)
.swapaxes(1,2)
.reshape(h, w))
For example,
c = np.arange(24).reshape((4,6))
print(c)
# [[ 0 1 2 3 4 5]
# [ 6 7 8 9 10 11]
# [12 13 14 15 16 17]
# [18 19 20 21 22 23]]
print(blockshaped(c, 2, 3))
# [[[ 0 1 2]
# [ 6 7 8]]
# [[ 3 4 5]
# [ 9 10 11]]
# [[12 13 14]
# [18 19 20]]
# [[15 16 17]
# [21 22 23]]]
print(unblockshaped(blockshaped(c, 2, 3), 4, 6))
# [[ 0 1 2 3 4 5]
# [ 6 7 8 9 10 11]
# [12 13 14 15 16 17]
# [18 19 20 21 22 23]]
Note that there is also superbatfish’s
blockwise_view
. It arranges the
blocks in a different format (using more axes) but it has the advantage of (1)
always returning a view and (2) being capable of handing arrays of any
dimension.