MAINT: skel3d: do not copy more than necessary

skimage/morphology/_skeletonize_3d.py

@@ -50,19 +50,19 @@ def skeletonize_3d(img):
     if img.ndim < 2 or img.ndim > 3:
         raise ValueError('expect 2D, got ndim = %s' % img.ndim)
 
-    img = img_as_ubyte(img)
     img = np.ascontiguousarray(img)
+    img = img_as_ubyte(img, force_copy=False)
 
-    img = img.copy()
+    # make an in image 3D pad w/ zeros to simplify dealing w/ boundaries
+    # NB: careful to not clobber the original *and* minimize copying
     if img.ndim == 2:
-        img = img[None, ...]
+        img_o = np.pad(img[None, ...], pad_width=1, mode='constant')
+    else:
+        img_o = np.pad(img, pad_width=1, mode='constant')
 
     # normalize to binary
-    maxval = img.max()
-    img[img != 0] = 1
-
-    # pad w/ zeros to simplify dealing w/ boundaries
-    img_o = np.pad(img, pad_width=1, mode='constant')
+    maxval = img_o.max()
+    img_o[img_o != 0] = 1
 
     # do the computation
     img_o = np.asarray(_compute_thin_image(img_o))

skimage/morphology/tests/test_skeletonize_3d.py

@@ -76,6 +76,25 @@ def test_dtype_conv():
                  img_as_ubyte(img).max())    # the intensity range is preserved
 
 
+def test_input():
+    # check that the input is not clobbered
+    # for 2D and 3D images of varying dtypes
+
+    imgs = [np.ones((8, 8), dtype=float), np.ones((4, 8, 8), dtype=float),
+            np.ones((8, 8), dtype=np.uint8), np.ones((4, 8, 8), dtype=np.uint8)]
+    for img in imgs:
+        yield check_input, img
+
+
+def check_input(img):
+    orig = img.copy()
+    with warnings.catch_warnings():
+        # UserWarning for possible precision loss, expected
+        warnings.simplefilter('ignore', UserWarning)
+        res = skeletonize_3d(img)
+    assert_equal(img, orig)
+
+
 def test_skeletonize_num_neighbours():
     # an empty image
     image = np.zeros((300, 300))