Merging non-overlapping array blocks
Use transpose/swapaxes to swap the second and third axes and then reshape to have the last two axes merged -
B.transpose(0,2,1,3).reshape(-1,B.shape[1]*B.shape[3])B.swapaxes(1,2).reshape(-1,B.shape[1]*B.shape[3])Sample run -
In [41]: AOut[41]: array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11], [12, 13, 14, 15]])In [42]: B = view_as_blocks(A, block_shape=(2, 2))In [43]: BOut[43]: array([[[[ 0, 1], [ 4, 5]], [[ 2, 3], [ 6, 7]]], [[[ 8, 9], [12, 13]], [[10, 11], [14, 15]]]])In [44]: B.transpose(0,2,1,3).reshape(-1,B.shape[1]*B.shape[3])Out[44]: array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11], [12, 13, 14, 15]])
This is where you'd better use einops:
from einops import rearrange# that's how you could rewrite view_as_blocksB = rearrange(A, '(x dx) (y dy) -> x y dx dy', dx=2, dy=2)# that's an answer to your questionA = rearrange(B, 'x y dx dy -> (x dx) (y dy)')See documentation for more operations on images