view-as-windows

view_as_windows

skimage.util.view_as_windows(arr_in, window_shape, step=1) [source]

Rolling window view of the input n-dimensional array.

Windows are overlapping views of the input array, with adjacent windows shifted by a single row or column (or an index of a higher dimension).

Parameters:

arr_in : ndarray N-d input array. window_shape : integer or tuple of length arr_in.ndim Defines the shape of the elementary n-dimensional orthotope (better know as hyperrectangle [R383]) of the rolling window view. If an integer is given, the shape will be a hypercube of sidelength given by its value. step : integer or tuple of length arr_in.ndim Indicates step size at which extraction shall be performed. If integer is given, then the step is uniform in all dimensions.

Returns:

arr_out : ndarray (rolling) window view of the input array. If arr_in is non-contiguous, a copy is made.

Notes

One should be very careful with rolling views when it comes to memory usage. Indeed, although a ‘view’ has the same memory footprint as its base array, the actual array that emerges when this ‘view’ is used in a computation is generally a (much) larger array than the original, especially for 2-dimensional arrays and above.

For example, let us consider a 3 dimensional array of size (100, 100, 100) of float64. This array takes about 8*100**3 Bytes for storage which is just 8 MB. If one decides to build a rolling view on this array with a window of (3, 3, 3) the hypothetical size of the rolling view (if one was to reshape the view for example) would be 8*(100-3+1)**3*3**3 which is about 203 MB! The scaling becomes even worse as the dimension of the input array becomes larger.

References

[R383]

(1, 2) http://en.wikipedia.org/wiki/Hyperrectangle

Examples

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

>>> import numpy as np >>> from skimage.util.shape import view_as_windows >>> A = np.arange(4*4).reshape(4,4) >>> A array([[ 0,  1,  2,  3],        [ 4,  5,  6,  7],        [ 8,  9, 10, 11],        [12, 13, 14, 15]]) >>> window_shape = (2, 2) >>> B = view_as_windows(A, window_shape) >>> B[0, 0] array([[0, 1],        [4, 5]]) >>> B[0, 1] array([[1, 2],        [5, 6]])

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

>>> A = np.arange(10) >>> A array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9]) >>> window_shape = (3,) >>> B = view_as_windows(A, window_shape) >>> B.shape (8, 3) >>> B array([[0, 1, 2],        [1, 2, 3],        [2, 3, 4],        [3, 4, 5],        [4, 5, 6],        [5, 6, 7],        [6, 7, 8],        [7, 8, 9]])

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

>>> A = np.arange(5*4).reshape(5, 4) >>> A array([[ 0,  1,  2,  3],        [ 4,  5,  6,  7],        [ 8,  9, 10, 11],        [12, 13, 14, 15],        [16, 17, 18, 19]]) >>> window_shape = (4, 3) >>> B = view_as_windows(A, window_shape) >>> B.shape (2, 2, 4, 3) >>> B  array([[[[ 0,  1,  2],          [ 4,  5,  6],          [ 8,  9, 10],          [12, 13, 14]],         [[ 1,  2,  3],          [ 5,  6,  7],          [ 9, 10, 11],          [13, 14, 15]]],        [[[ 4,  5,  6],          [ 8,  9, 10],          [12, 13, 14],          [16, 17, 18]],         [[ 5,  6,  7],          [ 9, 10, 11],          [13, 14, 15],          [17, 18, 19]]]])