Merge 9d7b12e into 7cc6af5

.travis.yml

@@ -21,6 +21,7 @@ jobs:
         - brew update
         - brew install openslide
         - echo 'export PATH="/usr/local/Frameworks/Python.framework/Versions/Current/bin:$PATH"' >> /Users/travis/.bash_profile
+        - source ~/.bash_profile
         - echo $PATH
       env: CAN_FAIL=true
     - name: "Python 3.7.4 on macOS"
@@ -30,6 +31,7 @@ jobs:
       before_install:
         - brew install openslide
         - echo 'export PATH="/usr/local/Frameworks/Python.framework/Versions/Current/bin:$PATH"' >> /Users/travis/.bash_profile
+        - source ~/.bash_profile
         - echo $PATH
     - name: "Python 3.8.3 on macOS"
       os: osx
@@ -38,6 +40,7 @@ jobs:
       before_install:
         - brew install openslide
         - echo 'export PATH="/usr/local/Frameworks/Python.framework/Versions/Current/bin:$PATH"' >> /Users/travis/.bash_profile
+        - source ~/.bash_profile
         - echo $PATH
     - name: "Python 3.6.8 on Windows"
       os: windows           # Windows 10.0.17134 N/A Build 17134
@@ -69,16 +72,18 @@ jobs:
         - wget https://github.com/openslide/openslide-winbuild/releases/download/v20171122/openslide-win64-20171122.zip -P /c/downloads
         - 7z e /c/downloads/openslide-win64-20171122.zip -aoa
         - export "PATH=/c/Python38/Scripts:/c/downloads/openslide-win64-20171122/bin:$PATH"
+        - echo alias python3=\'python\' >> ~.bashrc
+        - source ~/.bashrc
       env: CAN_FAIL=true
   allow_failures:
     - env: CAN_FAIL=true
   fast_finish: true
 
 # command to install dependencies
 install:
-  - pip3 install -e .[testing]
-  - pip3 install flake8
-  - pip3 install pooch
+  - python3 -m pip install -e .[testing]
+  - python3 -m pip install flake8
+  - python3 -m pip install pooch
 
 services:
   - xvfb

src/histolab/filters/morphological_filters.py

@@ -57,9 +57,9 @@ def __init__(
         self.avoid_overmask = avoid_overmask
         self.overmask_thresh = overmask_thresh
 
-    def __call__(self, np_img) -> np.ndarray:
+    def __call__(self, np_mask: np.ndarray) -> np.ndarray:
         return F.remove_small_objects(
-            np_img, self.min_size, self.avoid_overmask, self.overmask_thresh
+            np_mask, self.min_size, self.avoid_overmask, self.overmask_thresh
         )
 
     def __repr__(self):
@@ -85,7 +85,7 @@ class RemoveSmallHoles:
     def __init__(self, area_threshold: int = 3000):
         self.area_threshold = area_threshold
 
-    def __call__(self, np_mask) -> np.ndarray:
+    def __call__(self, np_mask: np.ndarray) -> np.ndarray:
         return skimage.morphology.remove_small_holes(np_mask, self.area_threshold)
 
     def __repr__(self):
@@ -208,7 +208,7 @@ def __init__(self, disk_size: int = 3, iterations: int = 1):
         self.disk_size = disk_size
         self.iterations = iterations
 
-    def __call__(self, np_mask) -> np.ndarray:
+    def __call__(self, np_mask: np.ndarray) -> np.ndarray:
         if not np.array_equal(np_mask, np_mask.astype(bool)):
             raise ValueError("Mask must be binary")
         return scipy.ndimage.morphology.binary_opening(
@@ -244,7 +244,7 @@ def __init__(self, disk_size: int = 3, iterations: int = 1):
         self.disk_size = disk_size
         self.iterations = iterations
 
-    def __call__(self, np_mask) -> np.ndarray:
+    def __call__(self, np_mask: np.ndarray) -> np.ndarray:
         if not np.array_equal(np_mask, np_mask.astype(bool)):
             raise ValueError("Mask must be binary")
         return scipy.ndimage.morphology.binary_closing(
@@ -255,6 +255,40 @@ def __repr__(self):
         return self.__class__.__name__ + "()"
 
 
+class WatershedSegmentation:
+    """Segment and label an binary mask with Watershed segmentation [1]_
+
+    The watershed algorithm treats pixels values as a local topography (elevation).
+
+    Parameters
+    ----------
+    np_mask : np.ndarray
+        Input mask
+    region_shape : int, optional
+        The local region within which to search for image peaks is defined as a squared
+        area region_shape x region_shape. Default is 6.
+
+    Returns
+    -------
+    np.ndarray
+        Labelled segmentation mask
+
+    References
+    --------
+    .. [1] Watershed segmentation.
+       https://scikit-image.org/docs/dev/auto_examples/segmentation/plot_watershed.html
+    """
+
+    def __init__(self, region_shape: int = 6) -> None:
+        self.region_shape = region_shape
+
+    def __call__(self, np_mask: np.ndarray) -> np.ndarray:
+        return F.watershed_segmentation(np_mask, self.region_shape)
+
+    def __repr__(self) -> str:
+        return self.__class__.__name__ + "()"
+
+
 class WhiteTopHat:
     """Return white top hat of an image.
 

src/histolab/filters/morphological_filters_functional.py

@@ -17,7 +17,10 @@
 # ------------------------------------------------------------------------
 
 import numpy as np
+import scipy.ndimage as sc_ndimage
+import skimage.feature as sk_feature
 import skimage.morphology as sk_morphology
+import skimage.segmentation as sk_segmentation
 
 from .util import mask_percent
 
@@ -60,3 +63,38 @@ def remove_small_objects(
             np_mask, new_min_size, avoid_overmask, overmask_thresh
         )
     return mask_no_small_object
+
+
+def watershed_segmentation(np_mask: np.ndarray, region_shape: int = 6) -> np.ndarray:
+    """Segment and label an binary mask with Watershed segmentation [1]_
+
+    The watershed algorithm treats pixels values as a local topography (elevation).
+
+    Parameters
+    ----------
+    np_mask : np.ndarray
+        Input mask
+    region_shape : int, optional
+        The local region within which to search for image peaks is defined as a squared
+        area region_shape x region_shape. Default is 6.
+
+    Returns
+    -------
+    np.ndarray
+        Labelled segmentation mask
+
+    References
+    --------
+    .. [1] Watershed segmentation.
+       https://scikit-image.org/docs/dev/auto_examples/segmentation/plot_watershed.html
+    """
+    distance = sc_ndimage.distance_transform_edt(np_mask)
+    local_maxi = sk_feature.peak_local_max(
+        distance,
+        indices=False,
+        footprint=np.ones((region_shape, region_shape)),
+        labels=np_mask,
+    )
+    markers = sc_ndimage.label(local_maxi)[0]
+    labels = sk_segmentation.watershed(-distance, markers, mask=np_mask)
+    return labels

tests/integration/test_morphological_filters.py

@@ -54,3 +54,21 @@ def test_remove_small_objects_filter(
 
     np.testing.assert_array_equal(mask_no_small_object, expected_value)
     assert type(mask_no_small_object) == np.ndarray
+
+
+@pytest.mark.parametrize(
+    "mask_array, region_shape, expected_array",
+    (
+        (MASKNPY.YTMA1, 6, "mask-arrays/ytma1-watershed-segmentation-region6"),
+        (MASKNPY.YTMA2, 6, "mask-arrays/ytma2-watershed-segmentation-region6"),
+        (MASKNPY.YTMA1, 3, "mask-arrays/ytma1-watershed-segmentation-region3"),
+        (MASKNPY.YTMA2, 3, "mask-arrays/ytma2-watershed-segmentation-region3"),
+    ),
+)
+def test_watershed_segmentation_filter(mask_array, region_shape, expected_array):
+    expected_value = load_expectation(expected_array, type_="npy")
+
+    mask_watershed = mof.watershed_segmentation(mask_array, region_shape)
+
+    np.testing.assert_array_equal(mask_watershed, expected_value)
+    assert type(mask_watershed) == np.ndarray

tests/unit/filters/test_morphological_fillters.py → tests/unit/filters/test_morphological_filters.py

@@ -1,9 +1,9 @@
 # encoding: utf-8
 
-import pytest
-
 import numpy as np
+import pytest
 import skimage.morphology
+
 from histolab.filters import morphological_filters as mof
 
 from ...base import IMAGE1_RGB, IMAGE2_RGBA
@@ -179,3 +179,17 @@ def it_calls_white_top_hat_filter(self, request):
         np.testing.assert_array_equal(_white_top_hat.call_args_list[0][0][0], mask_arr)
         np.testing.assert_array_equal(_white_top_hat.call_args_list[0][0][1], disk)
         assert type(white_top_hat(mask_arr)) == np.ndarray
+
+    def it_calls_watershed_segmentation_functional(self, request):
+        mask_arr = NpArrayMock.ONES_500X500X4_BOOL
+        F_watershed_segmentation = function_mock(
+            request,
+            "histolab.filters.morphological_filters_functional.watershed_segmentation",
+        )
+        F_watershed_segmentation.return_value = mask_arr
+        watershed_segmentation = mof.WatershedSegmentation()
+
+        watershed_segmentation(mask_arr)
+
+        F_watershed_segmentation.assert_called_once_with(mask_arr, 6)
+        assert type(watershed_segmentation(mask_arr)) == np.ndarray