From 733bf13297bc0c10ac8c5e6a74f0348f2f2436d7 Mon Sep 17 00:00:00 2001
From: Richard Brown <33289025+rijobro@users.noreply.github.com>
Date: Tue, 24 Aug 2021 15:21:23 +0100
Subject: [PATCH 1/3] all close

Signed-off-by: Richard Brown <33289025+rijobro@users.noreply.github.com>
---
 tests/test_fill_holes.py                       |  3 +--
 tests/test_flip.py                             |  7 ++-----
 tests/test_flipd.py                            |  7 ++-----
 tests/test_keep_largest_connected_component.py |  4 ++--
 tests/test_label_filter.py                     |  3 +--
 tests/test_rand_axis_flip.py                   |  8 ++------
 tests/test_rand_axis_flipd.py                  |  7 ++-----
 tests/test_rand_flip.py                        |  7 ++-----
 tests/test_rand_flipd.py                       |  7 ++-----
 tests/utils.py                                 | 17 +++++++----------
 10 files changed, 23 insertions(+), 47 deletions(-)

diff --git a/tests/test_fill_holes.py b/tests/test_fill_holes.py
index 294bbd8c87..dfc6fb5154 100644
--- a/tests/test_fill_holes.py
+++ b/tests/test_fill_holes.py
@@ -278,10 +278,9 @@ def test_correct_results(self, _, args, input_image, expected):
         converter = FillHoles(**args)
         if isinstance(input_image, torch.Tensor) and torch.cuda.is_available():
             result = converter(clone(input_image).cuda())
-            assert allclose(result, expected.cuda())
         else:
             result = converter(clone(input_image))
-            assert allclose(result, expected)
+        self.assertTrue(allclose(result, expected))
 
     @parameterized.expand(INVALID_CASES)
     def test_raise_exception(self, _, args, input_image, expected_error):
diff --git a/tests/test_flip.py b/tests/test_flip.py
index bd0162fb8b..f9dbb5368e 100644
--- a/tests/test_flip.py
+++ b/tests/test_flip.py
@@ -12,11 +12,10 @@
 import unittest
 
 import numpy as np
-import torch
 from parameterized import parameterized
 
 from monai.transforms import Flip
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
 
 INVALID_CASES = [("wrong_axis", ["s", 1], TypeError), ("not_numbers", "s", TypeError)]
 
@@ -40,9 +39,7 @@ def test_correct_results(self, _, spatial_axis):
                 expected.append(np.flip(channel, spatial_axis))
             expected = np.stack(expected)
             result = flip(im)
-            if isinstance(result, torch.Tensor):
-                result = result.cpu()
-            self.assertTrue(np.allclose(expected, result))
+            self.assertTrue(allclose(expected, result))
 
 
 if __name__ == "__main__":
diff --git a/tests/test_flipd.py b/tests/test_flipd.py
index cec4a99cbf..5266e36955 100644
--- a/tests/test_flipd.py
+++ b/tests/test_flipd.py
@@ -12,11 +12,10 @@
 import unittest
 
 import numpy as np
-import torch
 from parameterized import parameterized
 
 from monai.transforms import Flipd
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
 
 INVALID_CASES = [("wrong_axis", ["s", 1], TypeError), ("not_numbers", "s", TypeError)]
 
@@ -39,9 +38,7 @@ def test_correct_results(self, _, spatial_axis):
                 expected.append(np.flip(channel, spatial_axis))
             expected = np.stack(expected)
             result = flip({"img": p(self.imt[0])})["img"]
-            if isinstance(result, torch.Tensor):
-                result = result.cpu()
-            assert np.allclose(expected, result)
+            self.assertTrue(allclose(expected, result))
 
 
 if __name__ == "__main__":
diff --git a/tests/test_keep_largest_connected_component.py b/tests/test_keep_largest_connected_component.py
index 670dd2d2ee..91a2d79718 100644
--- a/tests/test_keep_largest_connected_component.py
+++ b/tests/test_keep_largest_connected_component.py
@@ -327,10 +327,10 @@ def test_correct_results(self, _, args, input_image, expected):
         converter = KeepLargestConnectedComponent(**args)
         if isinstance(input_image, torch.Tensor) and torch.cuda.is_available():
             result = converter(clone(input_image).cuda())
-            assert allclose(result, expected.cuda())
+
         else:
             result = converter(clone(input_image))
-            assert allclose(result, expected)
+        self.assertTrue(allclose(result, expected.cuda()))
 
     @parameterized.expand(INVALID_CASES)
     def test_raise_exception(self, _, args, input_image, expected_error):
diff --git a/tests/test_label_filter.py b/tests/test_label_filter.py
index 9165fddc40..87da61acc9 100644
--- a/tests/test_label_filter.py
+++ b/tests/test_label_filter.py
@@ -108,10 +108,9 @@ def test_correct_results(self, _, args, input_image, expected):
         converter = LabelFilter(**args)
         if isinstance(input_image, torch.Tensor) and torch.cuda.is_available():
             result = converter(clone(input_image).cuda())
-            assert allclose(result, expected.cuda())
         else:
             result = converter(clone(input_image))
-            assert allclose(result, expected)
+        self.assertTrue(allclose(result, expected))
 
     @parameterized.expand(INVALID_CASES)
     def test_raise_exception(self, _, args, input_image, expected_error):
diff --git a/tests/test_rand_axis_flip.py b/tests/test_rand_axis_flip.py
index bd53fa1fb0..9acd7be74c 100644
--- a/tests/test_rand_axis_flip.py
+++ b/tests/test_rand_axis_flip.py
@@ -12,10 +12,9 @@
 import unittest
 
 import numpy as np
-import torch
 
 from monai.transforms import RandAxisFlip
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
 
 
 class TestRandAxisFlip(NumpyImageTestCase2D):
@@ -23,13 +22,10 @@ def test_correct_results(self):
         for p in TEST_NDARRAYS:
             flip = RandAxisFlip(prob=1.0)
             result = flip(p(self.imt[0]))
-            if isinstance(result, torch.Tensor):
-                result = result.cpu()
-
             expected = []
             for channel in self.imt[0]:
                 expected.append(np.flip(channel, flip._axis))
-            self.assertTrue(np.allclose(np.stack(expected), result))
+            self.assertTrue(allclose(np.stack(expected), result))
 
 
 if __name__ == "__main__":
diff --git a/tests/test_rand_axis_flipd.py b/tests/test_rand_axis_flipd.py
index 518d78dd29..bba7eb632c 100644
--- a/tests/test_rand_axis_flipd.py
+++ b/tests/test_rand_axis_flipd.py
@@ -12,10 +12,9 @@
 import unittest
 
 import numpy as np
-import torch
 
 from monai.transforms import RandAxisFlipd
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase3D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase3D, allclose
 
 
 class TestRandAxisFlip(NumpyImageTestCase3D):
@@ -23,13 +22,11 @@ def test_correct_results(self):
         for p in TEST_NDARRAYS:
             flip = RandAxisFlipd(keys="img", prob=1.0)
             result = flip({"img": p(self.imt[0])})["img"]
-            if isinstance(result, torch.Tensor):
-                result = result.cpu()
 
             expected = []
             for channel in self.imt[0]:
                 expected.append(np.flip(channel, flip._axis))
-            self.assertTrue(np.allclose(np.stack(expected), result))
+            self.assertTrue(allclose(np.stack(expected), result))
 
 
 if __name__ == "__main__":
diff --git a/tests/test_rand_flip.py b/tests/test_rand_flip.py
index c20c13fec5..1d5cc34316 100644
--- a/tests/test_rand_flip.py
+++ b/tests/test_rand_flip.py
@@ -12,11 +12,10 @@
 import unittest
 
 import numpy as np
-import torch
 from parameterized import parameterized
 
 from monai.transforms import RandFlip
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
 
 INVALID_CASES = [("wrong_axis", ["s", 1], TypeError), ("not_numbers", "s", TypeError)]
 
@@ -40,9 +39,7 @@ def test_correct_results(self, _, spatial_axis):
                 expected.append(np.flip(channel, spatial_axis))
             expected = np.stack(expected)
             result = flip(im)
-            if isinstance(result, torch.Tensor):
-                result = result.cpu()
-            self.assertTrue(np.allclose(expected, result))
+            self.assertTrue(allclose(expected, result))
 
 
 if __name__ == "__main__":
diff --git a/tests/test_rand_flipd.py b/tests/test_rand_flipd.py
index 42c7dfe4b5..743c5a49e8 100644
--- a/tests/test_rand_flipd.py
+++ b/tests/test_rand_flipd.py
@@ -12,11 +12,10 @@
 import unittest
 
 import numpy as np
-import torch
 from parameterized import parameterized
 
 from monai.transforms import RandFlipd
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
 
 VALID_CASES = [("no_axis", None), ("one_axis", 1), ("many_axis", [0, 1])]
 
@@ -27,13 +26,11 @@ def test_correct_results(self, _, spatial_axis):
         for p in TEST_NDARRAYS:
             flip = RandFlipd(keys="img", prob=1.0, spatial_axis=spatial_axis)
             result = flip({"img": p(self.imt[0])})["img"]
-            if isinstance(result, torch.Tensor):
-                result = result.cpu()
             expected = []
             for channel in self.imt[0]:
                 expected.append(np.flip(channel, spatial_axis))
             expected = np.stack(expected)
-            self.assertTrue(np.allclose(expected, result))
+            self.assertTrue(allclose(expected, result))
 
 
 if __name__ == "__main__":
diff --git a/tests/utils.py b/tests/utils.py
index 1148af7551..1b2b7dea9b 100644
--- a/tests/utils.py
+++ b/tests/utils.py
@@ -33,6 +33,7 @@
 
 from monai.config import NdarrayTensor
 from monai.config.deviceconfig import USE_COMPILED
+from monai.config.type_definitions import NdarrayOrTensor
 from monai.data import create_test_image_2d, create_test_image_3d
 from monai.utils import ensure_tuple, optional_import, set_determinism
 from monai.utils.module import version_leq
@@ -55,7 +56,7 @@ def clone(data: NdarrayTensor) -> NdarrayTensor:
     return copy.deepcopy(data)
 
 
-def allclose(a: NdarrayTensor, b: NdarrayTensor) -> bool:
+def allclose(a: NdarrayOrTensor, b: NdarrayOrTensor, rtol=1.0e-5, atol=1.0e-8, equal_nan=False) -> bool:
     """
     Check if all values of two data objects are close.
 
@@ -63,19 +64,15 @@ def allclose(a: NdarrayTensor, b: NdarrayTensor) -> bool:
         This method also checks that both data objects are either Pytorch Tensors or numpy arrays.
 
     Args:
-        a (NdarrayTensor): Pytorch Tensor or numpy array for comparison
-        b (NdarrayTensor): Pytorch Tensor or numpy array to compare against
+        a (NdarrayOrTensor): Pytorch Tensor or numpy array for comparison
+        b (NdarrayOrTensor): Pytorch Tensor or numpy array to compare against
 
     Returns:
         bool: If both data objects are close.
     """
-    if isinstance(a, np.ndarray) and isinstance(b, np.ndarray):
-        return np.allclose(a, b)
-
-    if isinstance(a, torch.Tensor) and isinstance(b, torch.Tensor):
-        return torch.allclose(a, b)
-
-    return False
+    a = a.cpu() if isinstance(a, torch.Tensor) else a
+    b = b.cpu() if isinstance(b, torch.Tensor) else b
+    return np.allclose(a, b, rtol, atol, equal_nan)
 
 
 def test_pretrained_networks(network, input_param, device):

From 00e9f22da43cc7e667d929b3fa87968b774887e0 Mon Sep 17 00:00:00 2001
From: Richard Brown <33289025+rijobro@users.noreply.github.com>
Date: Tue, 24 Aug 2021 15:48:55 +0100
Subject: [PATCH 2/3] assert_allclose

Signed-off-by: Richard Brown <33289025+rijobro@users.noreply.github.com>
---
 tests/test_fill_holes.py                       |  4 ++--
 tests/test_flip.py                             |  4 ++--
 tests/test_flipd.py                            |  4 ++--
 tests/test_keep_largest_connected_component.py |  4 ++--
 tests/test_label_filter.py                     |  4 ++--
 tests/test_rand_axis_flip.py                   |  4 ++--
 tests/test_rand_axis_flipd.py                  |  4 ++--
 tests/test_rand_flip.py                        |  4 ++--
 tests/test_rand_flipd.py                       |  4 ++--
 tests/utils.py                                 | 12 +++---------
 10 files changed, 21 insertions(+), 27 deletions(-)

diff --git a/tests/test_fill_holes.py b/tests/test_fill_holes.py
index dfc6fb5154..6ea83c239b 100644
--- a/tests/test_fill_holes.py
+++ b/tests/test_fill_holes.py
@@ -16,7 +16,7 @@
 from parameterized import parameterized
 
 from monai.transforms import FillHoles
-from tests.utils import allclose, clone
+from tests.utils import assert_allclose, clone
 
 grid_1_raw = [
     [1, 1, 1],
@@ -280,7 +280,7 @@ def test_correct_results(self, _, args, input_image, expected):
             result = converter(clone(input_image).cuda())
         else:
             result = converter(clone(input_image))
-        self.assertTrue(allclose(result, expected))
+        assert_allclose(result, expected)
 
     @parameterized.expand(INVALID_CASES)
     def test_raise_exception(self, _, args, input_image, expected_error):
diff --git a/tests/test_flip.py b/tests/test_flip.py
index f9dbb5368e..404a3def7d 100644
--- a/tests/test_flip.py
+++ b/tests/test_flip.py
@@ -15,7 +15,7 @@
 from parameterized import parameterized
 
 from monai.transforms import Flip
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 INVALID_CASES = [("wrong_axis", ["s", 1], TypeError), ("not_numbers", "s", TypeError)]
 
@@ -39,7 +39,7 @@ def test_correct_results(self, _, spatial_axis):
                 expected.append(np.flip(channel, spatial_axis))
             expected = np.stack(expected)
             result = flip(im)
-            self.assertTrue(allclose(expected, result))
+            assert_allclose(expected, result)
 
 
 if __name__ == "__main__":
diff --git a/tests/test_flipd.py b/tests/test_flipd.py
index 5266e36955..1676723800 100644
--- a/tests/test_flipd.py
+++ b/tests/test_flipd.py
@@ -15,7 +15,7 @@
 from parameterized import parameterized
 
 from monai.transforms import Flipd
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 INVALID_CASES = [("wrong_axis", ["s", 1], TypeError), ("not_numbers", "s", TypeError)]
 
@@ -38,7 +38,7 @@ def test_correct_results(self, _, spatial_axis):
                 expected.append(np.flip(channel, spatial_axis))
             expected = np.stack(expected)
             result = flip({"img": p(self.imt[0])})["img"]
-            self.assertTrue(allclose(expected, result))
+            assert_allclose(expected, result)
 
 
 if __name__ == "__main__":
diff --git a/tests/test_keep_largest_connected_component.py b/tests/test_keep_largest_connected_component.py
index 91a2d79718..527d986614 100644
--- a/tests/test_keep_largest_connected_component.py
+++ b/tests/test_keep_largest_connected_component.py
@@ -15,7 +15,7 @@
 from parameterized import parameterized
 
 from monai.transforms import KeepLargestConnectedComponent
-from tests.utils import allclose, clone
+from tests.utils import assert_allclose, clone
 
 grid_1 = torch.tensor([[[0, 0, 1, 0, 0], [0, 2, 1, 1, 1], [1, 2, 1, 0, 0], [1, 2, 0, 1, 0], [2, 2, 0, 0, 2]]])
 grid_2 = torch.tensor([[[0, 0, 0, 0, 1], [0, 0, 1, 1, 1], [1, 0, 1, 1, 2], [1, 0, 1, 2, 2], [0, 0, 0, 0, 1]]])
@@ -330,7 +330,7 @@ def test_correct_results(self, _, args, input_image, expected):
 
         else:
             result = converter(clone(input_image))
-        self.assertTrue(allclose(result, expected.cuda()))
+        assert_allclose(result, expected)
 
     @parameterized.expand(INVALID_CASES)
     def test_raise_exception(self, _, args, input_image, expected_error):
diff --git a/tests/test_label_filter.py b/tests/test_label_filter.py
index 87da61acc9..c699fb31fd 100644
--- a/tests/test_label_filter.py
+++ b/tests/test_label_filter.py
@@ -16,7 +16,7 @@
 from parameterized import parameterized
 
 from monai.transforms import LabelFilter
-from tests.utils import allclose, clone
+from tests.utils import assert_allclose, clone
 
 grid_1 = torch.tensor(
     [
@@ -110,7 +110,7 @@ def test_correct_results(self, _, args, input_image, expected):
             result = converter(clone(input_image).cuda())
         else:
             result = converter(clone(input_image))
-        self.assertTrue(allclose(result, expected))
+        assert_allclose(result, expected)
 
     @parameterized.expand(INVALID_CASES)
     def test_raise_exception(self, _, args, input_image, expected_error):
diff --git a/tests/test_rand_axis_flip.py b/tests/test_rand_axis_flip.py
index 9acd7be74c..c05c3a1e0d 100644
--- a/tests/test_rand_axis_flip.py
+++ b/tests/test_rand_axis_flip.py
@@ -14,7 +14,7 @@
 import numpy as np
 
 from monai.transforms import RandAxisFlip
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 
 class TestRandAxisFlip(NumpyImageTestCase2D):
@@ -25,7 +25,7 @@ def test_correct_results(self):
             expected = []
             for channel in self.imt[0]:
                 expected.append(np.flip(channel, flip._axis))
-            self.assertTrue(allclose(np.stack(expected), result))
+            assert_allclose(np.stack(expected), result)
 
 
 if __name__ == "__main__":
diff --git a/tests/test_rand_axis_flipd.py b/tests/test_rand_axis_flipd.py
index bba7eb632c..7bef0baa63 100644
--- a/tests/test_rand_axis_flipd.py
+++ b/tests/test_rand_axis_flipd.py
@@ -14,7 +14,7 @@
 import numpy as np
 
 from monai.transforms import RandAxisFlipd
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase3D, allclose
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase3D, assert_allclose
 
 
 class TestRandAxisFlip(NumpyImageTestCase3D):
@@ -26,7 +26,7 @@ def test_correct_results(self):
             expected = []
             for channel in self.imt[0]:
                 expected.append(np.flip(channel, flip._axis))
-            self.assertTrue(allclose(np.stack(expected), result))
+            assert_allclose(np.stack(expected), result)
 
 
 if __name__ == "__main__":
diff --git a/tests/test_rand_flip.py b/tests/test_rand_flip.py
index 1d5cc34316..b3c514cb1f 100644
--- a/tests/test_rand_flip.py
+++ b/tests/test_rand_flip.py
@@ -15,7 +15,7 @@
 from parameterized import parameterized
 
 from monai.transforms import RandFlip
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 INVALID_CASES = [("wrong_axis", ["s", 1], TypeError), ("not_numbers", "s", TypeError)]
 
@@ -39,7 +39,7 @@ def test_correct_results(self, _, spatial_axis):
                 expected.append(np.flip(channel, spatial_axis))
             expected = np.stack(expected)
             result = flip(im)
-            self.assertTrue(allclose(expected, result))
+            assert_allclose(expected, result)
 
 
 if __name__ == "__main__":
diff --git a/tests/test_rand_flipd.py b/tests/test_rand_flipd.py
index 743c5a49e8..8972024fd8 100644
--- a/tests/test_rand_flipd.py
+++ b/tests/test_rand_flipd.py
@@ -15,7 +15,7 @@
 from parameterized import parameterized
 
 from monai.transforms import RandFlipd
-from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, allclose
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 VALID_CASES = [("no_axis", None), ("one_axis", 1), ("many_axis", [0, 1])]
 
@@ -30,7 +30,7 @@ def test_correct_results(self, _, spatial_axis):
             for channel in self.imt[0]:
                 expected.append(np.flip(channel, spatial_axis))
             expected = np.stack(expected)
-            self.assertTrue(allclose(expected, result))
+            assert_allclose(expected, result)
 
 
 if __name__ == "__main__":
diff --git a/tests/utils.py b/tests/utils.py
index 1b2b7dea9b..22720849f1 100644
--- a/tests/utils.py
+++ b/tests/utils.py
@@ -56,23 +56,17 @@ def clone(data: NdarrayTensor) -> NdarrayTensor:
     return copy.deepcopy(data)
 
 
-def allclose(a: NdarrayOrTensor, b: NdarrayOrTensor, rtol=1.0e-5, atol=1.0e-8, equal_nan=False) -> bool:
+def assert_allclose(a: NdarrayOrTensor, b: NdarrayOrTensor, *args, **kwargs):
     """
-    Check if all values of two data objects are close.
-
-    Note:
-        This method also checks that both data objects are either Pytorch Tensors or numpy arrays.
+    Assert that all values of two data objects are close.
 
     Args:
         a (NdarrayOrTensor): Pytorch Tensor or numpy array for comparison
         b (NdarrayOrTensor): Pytorch Tensor or numpy array to compare against
-
-    Returns:
-        bool: If both data objects are close.
     """
     a = a.cpu() if isinstance(a, torch.Tensor) else a
     b = b.cpu() if isinstance(b, torch.Tensor) else b
-    return np.allclose(a, b, rtol, atol, equal_nan)
+    np.testing.assert_allclose(a, b, *args, **kwargs)
 
 
 def test_pretrained_networks(network, input_param, device):

From 586737936976a225a3eb5eaf20f360d2165d0f68 Mon Sep 17 00:00:00 2001
From: Richard Brown <33289025+rijobro@users.noreply.github.com>
Date: Tue, 24 Aug 2021 17:03:29 +0100
Subject: [PATCH 3/3] ScaleIntensity

Signed-off-by: Richard Brown <33289025+rijobro@users.noreply.github.com>
---
 monai/data/synthetic.py                  |  4 ++--
 monai/transforms/intensity/array.py      | 14 +++++++----
 monai/transforms/intensity/dictionary.py |  8 +++++--
 monai/transforms/utils.py                | 12 ++++++----
 monai/visualize/img2tensorboard.py       |  2 +-
 tests/test_rand_scale_intensity.py       | 15 ++++++------
 tests/test_rand_scale_intensityd.py      | 17 +++++++-------
 tests/test_scale_intensity.py            | 26 ++++++++++----------
 tests/test_scale_intensityd.py           | 30 +++++++++++++-----------
 9 files changed, 74 insertions(+), 54 deletions(-)

diff --git a/monai/data/synthetic.py b/monai/data/synthetic.py
index 20a7829cab..6eec9fd277 100644
--- a/monai/data/synthetic.py
+++ b/monai/data/synthetic.py
@@ -76,7 +76,7 @@ def create_test_image_2d(
     labels = np.ceil(image).astype(np.int32)
 
     norm = rs.uniform(0, num_seg_classes * noise_max, size=image.shape)
-    noisyimage = rescale_array(np.maximum(image, norm))
+    noisyimage: np.ndarray = rescale_array(np.maximum(image, norm))  # type: ignore
 
     if channel_dim is not None:
         if not (isinstance(channel_dim, int) and channel_dim in (-1, 0, 2)):
@@ -151,7 +151,7 @@ def create_test_image_3d(
     labels = np.ceil(image).astype(np.int32)
 
     norm = rs.uniform(0, num_seg_classes * noise_max, size=image.shape)
-    noisyimage = rescale_array(np.maximum(image, norm))
+    noisyimage: np.ndarray = rescale_array(np.maximum(image, norm))  # type: ignore
 
     if channel_dim is not None:
         if not (isinstance(channel_dim, int) and channel_dim in (-1, 0, 3)):
diff --git a/monai/transforms/intensity/array.py b/monai/transforms/intensity/array.py
index 4b7c8d6997..c2c4b3c38b 100644
--- a/monai/transforms/intensity/array.py
+++ b/monai/transforms/intensity/array.py
@@ -373,6 +373,8 @@ class ScaleIntensity(Transform):
     If `minv` and `maxv` not provided, use `factor` to scale image by ``v = v * (1 + factor)``.
     """
 
+    backend = [TransformBackends.TORCH, TransformBackends.NUMPY]
+
     def __init__(
         self, minv: Optional[float] = 0.0, maxv: Optional[float] = 1.0, factor: Optional[float] = None
     ) -> None:
@@ -387,7 +389,7 @@ def __init__(
         self.maxv = maxv
         self.factor = factor
 
-    def __call__(self, img: np.ndarray) -> np.ndarray:
+    def __call__(self, img: NdarrayOrTensor) -> NdarrayOrTensor:
         """
         Apply the transform to `img`.
 
@@ -396,9 +398,11 @@ def __call__(self, img: np.ndarray) -> np.ndarray:
 
         """
         if self.minv is not None and self.maxv is not None:
-            return np.asarray(rescale_array(img, self.minv, self.maxv, img.dtype))
+            return rescale_array(img, self.minv, self.maxv, img.dtype)
         if self.factor is not None:
-            return np.asarray(img * (1 + self.factor), dtype=img.dtype)
+            out = img * (1 + self.factor)
+            out, *_ = convert_data_type(out, dtype=img.dtype)
+            return out
         raise ValueError("Incompatible values: minv=None or maxv=None and factor=None.")
 
 
@@ -408,6 +412,8 @@ class RandScaleIntensity(RandomizableTransform):
     is randomly picked.
     """
 
+    backend = ScaleIntensity.backend
+
     def __init__(self, factors: Union[Tuple[float, float], float], prob: float = 0.1) -> None:
         """
         Args:
@@ -429,7 +435,7 @@ def randomize(self, data: Optional[Any] = None) -> None:
         self.factor = self.R.uniform(low=self.factors[0], high=self.factors[1])
         super().randomize(None)
 
-    def __call__(self, img: np.ndarray) -> np.ndarray:
+    def __call__(self, img: NdarrayOrTensor) -> NdarrayOrTensor:
         """
         Apply the transform to `img`.
         """
diff --git a/monai/transforms/intensity/dictionary.py b/monai/transforms/intensity/dictionary.py
index 522007df29..ad57353d07 100644
--- a/monai/transforms/intensity/dictionary.py
+++ b/monai/transforms/intensity/dictionary.py
@@ -472,6 +472,8 @@ class ScaleIntensityd(MapTransform):
     If `minv` and `maxv` not provided, use `factor` to scale image by ``v = v * (1 + factor)``.
     """
 
+    backend = ScaleIntensity.backend
+
     def __init__(
         self,
         keys: KeysCollection,
@@ -494,7 +496,7 @@ def __init__(
         super().__init__(keys, allow_missing_keys)
         self.scaler = ScaleIntensity(minv, maxv, factor)
 
-    def __call__(self, data: Mapping[Hashable, np.ndarray]) -> Dict[Hashable, np.ndarray]:
+    def __call__(self, data: Mapping[Hashable, NdarrayOrTensor]) -> Dict[Hashable, NdarrayOrTensor]:
         d = dict(data)
         for key in self.key_iterator(d):
             d[key] = self.scaler(d[key])
@@ -506,6 +508,8 @@ class RandScaleIntensityd(RandomizableTransform, MapTransform):
     Dictionary-based version :py:class:`monai.transforms.RandScaleIntensity`.
     """
 
+    backend = ScaleIntensity.backend
+
     def __init__(
         self,
         keys: KeysCollection,
@@ -539,7 +543,7 @@ def randomize(self, data: Optional[Any] = None) -> None:
         self.factor = self.R.uniform(low=self.factors[0], high=self.factors[1])
         super().randomize(None)
 
-    def __call__(self, data: Mapping[Hashable, np.ndarray]) -> Dict[Hashable, np.ndarray]:
+    def __call__(self, data: Mapping[Hashable, NdarrayOrTensor]) -> Dict[Hashable, NdarrayOrTensor]:
         d = dict(data)
         self.randomize()
         if not self._do_transform:
diff --git a/monai/transforms/utils.py b/monai/transforms/utils.py
index e81cb7ca17..e3e61b6c97 100644
--- a/monai/transforms/utils.py
+++ b/monai/transforms/utils.py
@@ -22,6 +22,7 @@
 import monai
 import monai.transforms.transform
 from monai.config import DtypeLike, IndexSelection
+from monai.config.type_definitions import NdarrayOrTensor
 from monai.networks.layers import GaussianFilter
 from monai.transforms.compose import Compose, OneOf
 from monai.transforms.transform import MapTransform, Transform
@@ -37,6 +38,7 @@
     min_version,
     optional_import,
 )
+from monai.utils.type_conversion import convert_data_type
 
 measure, _ = optional_import("skimage.measure", "0.14.2", min_version)
 ndimage, _ = optional_import("scipy.ndimage")
@@ -130,15 +132,17 @@ def zero_margins(img: np.ndarray, margin: int) -> bool:
     return not np.any(img[:, :margin, :]) and not np.any(img[:, -margin:, :])
 
 
-def rescale_array(arr: np.ndarray, minv: float = 0.0, maxv: float = 1.0, dtype: DtypeLike = np.float32):
+def rescale_array(
+    arr: NdarrayOrTensor, minv: float = 0.0, maxv: float = 1.0, dtype: Union[DtypeLike, torch.dtype] = np.float32
+) -> NdarrayOrTensor:
     """
     Rescale the values of numpy array `arr` to be from `minv` to `maxv`.
     """
     if dtype is not None:
-        arr = arr.astype(dtype)
+        arr, *_ = convert_data_type(arr, dtype=dtype)
 
-    mina = np.min(arr)
-    maxa = np.max(arr)
+    mina = arr.min()
+    maxa = arr.max()
 
     if mina == maxa:
         return arr * minv
diff --git a/monai/visualize/img2tensorboard.py b/monai/visualize/img2tensorboard.py
index 4a17607320..ccdbdc2396 100644
--- a/monai/visualize/img2tensorboard.py
+++ b/monai/visualize/img2tensorboard.py
@@ -188,7 +188,7 @@ def plot_2d_or_3d_image(
     d: np.ndarray = data_index.detach().cpu().numpy() if isinstance(data_index, torch.Tensor) else data_index
 
     if d.ndim == 2:
-        d = rescale_array(d, 0, 1)
+        d = rescale_array(d, 0, 1)  # type: ignore
         dataformats = "HW"
         writer.add_image(f"{tag}_{dataformats}", d, step, dataformats=dataformats)
         return
diff --git a/tests/test_rand_scale_intensity.py b/tests/test_rand_scale_intensity.py
index 2126301758..750d88bfad 100644
--- a/tests/test_rand_scale_intensity.py
+++ b/tests/test_rand_scale_intensity.py
@@ -14,17 +14,18 @@
 import numpy as np
 
 from monai.transforms import RandScaleIntensity
-from tests.utils import NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 
 class TestRandScaleIntensity(NumpyImageTestCase2D):
     def test_value(self):
-        scaler = RandScaleIntensity(factors=0.5, prob=1.0)
-        scaler.set_random_state(seed=0)
-        result = scaler(self.imt)
-        np.random.seed(0)
-        expected = (self.imt * (1 + np.random.uniform(low=-0.5, high=0.5))).astype(np.float32)
-        np.testing.assert_allclose(result, expected)
+        for p in TEST_NDARRAYS:
+            scaler = RandScaleIntensity(factors=0.5, prob=1.0)
+            scaler.set_random_state(seed=0)
+            result = scaler(p(self.imt))
+            np.random.seed(0)
+            expected = p((self.imt * (1 + np.random.uniform(low=-0.5, high=0.5))).astype(np.float32))
+            assert_allclose(result, expected, rtol=1e-7, atol=0)
 
 
 if __name__ == "__main__":
diff --git a/tests/test_rand_scale_intensityd.py b/tests/test_rand_scale_intensityd.py
index 6e207e3cc2..a8d2e63f65 100644
--- a/tests/test_rand_scale_intensityd.py
+++ b/tests/test_rand_scale_intensityd.py
@@ -14,18 +14,19 @@
 import numpy as np
 
 from monai.transforms import RandScaleIntensityd
-from tests.utils import NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 
 class TestRandScaleIntensityd(NumpyImageTestCase2D):
     def test_value(self):
-        key = "img"
-        scaler = RandScaleIntensityd(keys=[key], factors=0.5, prob=1.0)
-        scaler.set_random_state(seed=0)
-        result = scaler({key: self.imt})
-        np.random.seed(0)
-        expected = (self.imt * (1 + np.random.uniform(low=-0.5, high=0.5))).astype(np.float32)
-        np.testing.assert_allclose(result[key], expected)
+        for p in TEST_NDARRAYS:
+            key = "img"
+            scaler = RandScaleIntensityd(keys=[key], factors=0.5, prob=1.0)
+            scaler.set_random_state(seed=0)
+            result = scaler({key: p(self.imt)})
+            np.random.seed(0)
+            expected = (self.imt * (1 + np.random.uniform(low=-0.5, high=0.5))).astype(np.float32)
+            assert_allclose(result[key], expected)
 
 
 if __name__ == "__main__":
diff --git a/tests/test_scale_intensity.py b/tests/test_scale_intensity.py
index 61e89191fd..c2485af616 100644
--- a/tests/test_scale_intensity.py
+++ b/tests/test_scale_intensity.py
@@ -14,24 +14,26 @@
 import numpy as np
 
 from monai.transforms import ScaleIntensity
-from tests.utils import NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 
 class TestScaleIntensity(NumpyImageTestCase2D):
     def test_range_scale(self):
-        scaler = ScaleIntensity(minv=1.0, maxv=2.0)
-        result = scaler(self.imt)
-        mina = np.min(self.imt)
-        maxa = np.max(self.imt)
-        norm = (self.imt - mina) / (maxa - mina)
-        expected = (norm * (2.0 - 1.0)) + 1.0
-        np.testing.assert_allclose(result, expected)
+        for p in TEST_NDARRAYS:
+            scaler = ScaleIntensity(minv=1.0, maxv=2.0)
+            result = scaler(p(self.imt))
+            mina = self.imt.min()
+            maxa = self.imt.max()
+            norm = (self.imt - mina) / (maxa - mina)
+            expected = p((norm * (2.0 - 1.0)) + 1.0)
+            assert_allclose(result, expected, rtol=1e-7, atol=0)
 
     def test_factor_scale(self):
-        scaler = ScaleIntensity(minv=None, maxv=None, factor=0.1)
-        result = scaler(self.imt)
-        expected = (self.imt * (1 + 0.1)).astype(np.float32)
-        np.testing.assert_allclose(result, expected)
+        for p in TEST_NDARRAYS:
+            scaler = ScaleIntensity(minv=None, maxv=None, factor=0.1)
+            result = scaler(p(self.imt))
+            expected = p((self.imt * (1 + 0.1)).astype(np.float32))
+            assert_allclose(result, expected, rtol=1e-7, atol=0)
 
 
 if __name__ == "__main__":
diff --git a/tests/test_scale_intensityd.py b/tests/test_scale_intensityd.py
index 688c99c6af..6e13dbc272 100644
--- a/tests/test_scale_intensityd.py
+++ b/tests/test_scale_intensityd.py
@@ -14,26 +14,28 @@
 import numpy as np
 
 from monai.transforms import ScaleIntensityd
-from tests.utils import NumpyImageTestCase2D
+from tests.utils import TEST_NDARRAYS, NumpyImageTestCase2D, assert_allclose
 
 
 class TestScaleIntensityd(NumpyImageTestCase2D):
     def test_range_scale(self):
-        key = "img"
-        scaler = ScaleIntensityd(keys=[key], minv=1.0, maxv=2.0)
-        result = scaler({key: self.imt})
-        mina = np.min(self.imt)
-        maxa = np.max(self.imt)
-        norm = (self.imt - mina) / (maxa - mina)
-        expected = (norm * (2.0 - 1.0)) + 1.0
-        np.testing.assert_allclose(result[key], expected)
+        for p in TEST_NDARRAYS:
+            key = "img"
+            scaler = ScaleIntensityd(keys=[key], minv=1.0, maxv=2.0)
+            result = scaler({key: p(self.imt)})
+            mina = np.min(self.imt)
+            maxa = np.max(self.imt)
+            norm = (self.imt - mina) / (maxa - mina)
+            expected = (norm * (2.0 - 1.0)) + 1.0
+            assert_allclose(result[key], expected)
 
     def test_factor_scale(self):
-        key = "img"
-        scaler = ScaleIntensityd(keys=[key], minv=None, maxv=None, factor=0.1)
-        result = scaler({key: self.imt})
-        expected = (self.imt * (1 + 0.1)).astype(np.float32)
-        np.testing.assert_allclose(result[key], expected)
+        for p in TEST_NDARRAYS:
+            key = "img"
+            scaler = ScaleIntensityd(keys=[key], minv=None, maxv=None, factor=0.1)
+            result = scaler({key: p(self.imt)})
+            expected = (self.imt * (1 + 0.1)).astype(np.float32)
+            assert_allclose(result[key], expected)
 
 
 if __name__ == "__main__":