replace _discard_edges with skimage.util.arraypad.crop

skimage/measure/_structural_similarity.py

@@ -6,6 +6,7 @@ import numpy as np
 from scipy.ndimage.filters import uniform_filter, convolve1d
 
 from ..util.dtype import dtype_range
+from ..util.arraypad import crop
 
 
 def gaussian_filter2(X, sigma=1.5, size=11):
@@ -41,39 +42,6 @@ def gaussian_filter2(X, sigma=1.5, size=11):
     return X
 
 
-def _discard_edges(X, pad):
-    """ Remove border of width pad from ndarray X.
-
-    Parameters
-    ----------
-    X : ndarray
-        image
-    pad : int or sequence of ints
-        border width to remove.  Can be a list of values corresponding to each
-        axis.  If pad is an integer, the same width is removed from all axes.
-
-    Returns
-    -------
-    Y : nadarray
-        image with edges removed
-
-    """
-    X = np.asanyarray(X)
-    if pad == 0:
-        return X
-
-    if isinstance(pad, int):
-        slice_array = [slice(pad, -pad), ] * X.ndim
-    else:
-        if len(pad) != X.ndim:
-            raise ValueError("pad array must match number of X dimensions")
-        slice_array = []
-        for d in range(X.ndim):
-            slice_array.append(slice(pad[d], -pad[d]))
-
-    return X[slice_array]
-
-
 def structural_similarity(X, Y, win_size=None, gradient=False,
                           dynamic_range=None, multichannel=None,
                           gaussian_weights=False, full=False,
@@ -278,9 +246,9 @@ def structural_similarity(X, Y, win_size=None, gradient=False,
         # weight with Eq. 7 of Wang and Simoncelli 2006.
         W = np.log((1 + vx / C2) * (1 + vy / C2))
         W /= W.sum()
-        mssim = _discard_edges(S * W, pad).sum()
+        mssim = crop(S * W, pad).sum()
     else:
-        mssim = _discard_edges(S, pad).mean()
+        mssim = crop(S, pad).mean()
 
     if gradient:
         # The following is Eqs. 7-8 of Avanaki 2009.

skimage/measure/tests/test_structural_similarity.py

@@ -5,8 +5,7 @@ from numpy.testing import (assert_equal, assert_raises, assert_almost_equal,
                            assert_array_almost_equal)
 
 from skimage.measure import structural_similarity as ssim
-from skimage.measure._structural_similarity import (gaussian_filter2,
-                                                    _discard_edges)
+from skimage.measure._structural_similarity import (gaussian_filter2)
 import skimage.data
 from skimage.io import imread
 from skimage import data_dir
@@ -232,23 +231,5 @@ def test_gaussian_filter2():
     assert np.all(xf[:, :3] == 0)
 
 
-def test_discard_edges():
-    x = np.zeros((11, 11))
-    x[3:8, 3:8] = 1.0
-    xd = _discard_edges(x, 3)
-    assert xd.shape == (5, 5)
-    assert np.all(xd == 1.0)
-
-    # non-uniform edge case
-    x = np.zeros((11, 11))
-    x[3:8, 1:10] = 1.0
-    xd = _discard_edges(x, [3, 1])
-    assert xd.shape == (5, 9)
-    assert np.all(xd == 1.0)
-
-    assert_raises(ValueError, _discard_edges, x, [3, 3, 3])
-    assert_raises(TypeError, _discard_edges, x, 3.5)
-
-
 if __name__ == "__main__":
     np.testing.run_module_suite()