Skip to content

Instantly share code, notes, and snippets.



Last active Dec 22, 2015
What would you like to do?
A blocproc function for scikits-image
"""My thoughts on what a skimage version of Matlab's blockproc [1] function could look like.
def process_windows(arr_in, window_shape, proc_func=None, step_size=(0,0),
border_size=(0,0), border_fill_mode='overlap', pad_mode='edge',
"""Map image windows for (optional) parallel according to specified
windowing rule.
arr_in: ndarray
The 2-dimensional input array.
window_shape: tuple
The 2-dimensional shape of the rolling window view.
proc_func: function
This function will be called for each resulting window. The
function should be of the form:
function(arr_window, window_def)
where arr_window is the view of arr_in based on the window and
border as specified and window_def is a tuple with:
window_def = (window_shape, border_size, window_bbox)
where window_bbox is the 2-dimensional bounding box of the
resulting window.
step_size: tuple
Step size (2-dimensional). Defaults to window_shape so that
default behaviour gives non-overlapping windows.
border_size: tuple
The amount (V,H) of border pixels to add to each window. The
function adds V rows above and below each window and H columns
left and right of each window.
The size of each resulting window will be:
[M + 2*V, N + 2*H]
border_fill_mode: {str, function}
Can be 'overlap' or any mode as accepted by skimage.util.pad.
The 'overlap' mode will fill the border with the border values
from arr_in itself, while skimage.util.pad will pad with values
as defined by the pad mode. When 'overlap' is used, border value
coordinates that fall outside of arr_in will be padded according
to pad_mode.
pad_mode: {str, function}
Any mode that is accepted by skimage.util.pad. Will be used to
fill border if border overlap falls outside of arr_in.
arr_out: ndarray
(rolling) window view of the input array.
See skimage.util.view_as_windows.
return_values: tuple
The window result as returned by each window's proc_func.
>>> 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)
>>> proc_func = lambda arr_window, window_def: np.sum(arr_window)
>>> B, results = process_windows(A, window_shape, proc_func)
>>> B[0, 0]
array([[0, 1],
[4, 5]])
>>> results[0,0]
>>> B[0, 1]
array([[1, 2],
[5, 6]])
>>> results[0,1]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment