Merging non-overlapping array blocks

Question

I divided a (512x512) 2-dimensional array to 2x2 blocks using this function.

skimage.util.view_as_blocks (arr_in, block_shape)
array([[ 0,  1,  2,  3],
   [ 4,  5,  6,  7],
   [ 8,  9, 10, 11],
   [12, 13, 14, 15]])
   >>> B = view_as_blocks(A, block_shape=(2, 2))
   >>> B[0, 0]
   array([[0, 1],
          [4, 5]])
   >>> B[0, 1]
   array([[2, 3],
          [6, 7]])

Now I need to put the same blocks to their original places after manipulation but I couldn't see any function in skimage for that.

What's the best way to merge the non-overlapping arrays as it was before?

Thank you!


Show source
| numpy   | python   | skimage   2017-01-06 17:01 1 Answers

Answers ( 1 )

  1. 2017-01-06 17:01

    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]: A
    Out[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]: B
    Out[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]])
    
◀ Go back