Merge pull request pytroll#2521 from mraspaud/feature_median_filter

Add a median filter modifier

continuous_integration/environment.yaml

@@ -5,6 +5,7 @@ dependencies:
   - xarray!=2022.9.0
   - dask
   - distributed
+  - dask-image
   - donfig
   - appdirs
   - toolz

doc/rtd_environment.yml

@@ -6,6 +6,7 @@ dependencies:
   - pip
   - appdirs
   - dask
+  - dask-image
   - defusedxml
   - donfig
   # 2.19.1 seems to cause library linking issues

satpy/etc/composites/visir.yaml

@@ -96,6 +96,11 @@ modifiers:
     - solar_azimuth_angle
     - solar_zenith_angle
 
+  median5x5:
+    modifier: !!python/name:satpy.modifiers.filters.Median
+    median_filter_params:
+      size: 5
+
 composites:
 
   airmass:

satpy/modifiers/filters.py

@@ -0,0 +1,34 @@
+"""Tests for image filters."""
+import logging
+
+import xarray as xr
+
+from satpy.modifiers import ModifierBase
+
+logger = logging.getLogger(__name__)
+
+
+class Median(ModifierBase):
+    """Apply a median filter to the band."""
+
+    def __init__(self, median_filter_params, **kwargs):
+        """Create the instance.
+
+        Args:
+            median_filter_params: The arguments to pass to dask-image's median_filter function. For example, {size: 3}
+                                  makes give the median filter a kernel of size 3.
+
+        """
+        self.median_filter_params = median_filter_params
+        super().__init__(**kwargs)
+
+    def __call__(self, arrays, **info):
+        """Get the median filtered band."""
+        from dask_image.ndfilters import median_filter
+
+        data = arrays[0]
+        logger.debug(f"Apply median filtering with parameters {self.median_filter_params}.")
+        res = xr.DataArray(median_filter(data.data, **self.median_filter_params),
+                           dims=data.dims, attrs=data.attrs, coords=data.coords)
+        self.apply_modifier_info(data, res)
+        return res

satpy/tests/modifier_tests/test_filters.py

@@ -0,0 +1,35 @@
+"""Implementation of some image filters."""
+
+import logging
+
+import dask.array as da
+import numpy as np
+import xarray as xr
+
+from satpy.modifiers.filters import Median
+
+
+def test_median(caplog):
+    """Test the median filter modifier."""
+    caplog.set_level(logging.DEBUG)
+    dims = "y", "x"
+    coordinates = dict(x=np.arange(6), y=np.arange(6))
+    attrs = dict(units="K")
+    median_filter_params = dict(size=3)
+    name = "median_filter"
+    median_filter = Median(median_filter_params, name=name)
+    array = xr.DataArray(da.arange(36).reshape((6, 6)), coords=coordinates, dims=dims, attrs=attrs)
+    res = median_filter([array])
+    filtered_array = np.array([[1, 2, 3, 4, 5, 5],
+                               [6, 7, 8, 9, 10, 11],
+                               [12, 13, 14, 15, 16, 17],
+                               [18, 19, 20, 21, 22, 23],
+                               [24, 25, 26, 27, 28, 29],
+                               [30, 30, 31, 32, 33, 34]])
+    np.testing.assert_allclose(res, filtered_array)
+    assert res.dims == dims
+    assert attrs.items() <= res.attrs.items()
+    assert res.attrs["name"] == name
+    np.testing.assert_equal(res.coords["x"], coordinates["x"])
+    np.testing.assert_equal(res.coords["y"], coordinates["y"])
+    assert "Apply median filtering with parameters {'size': 3}" in caplog.text

setup.py

@@ -39,7 +39,7 @@
                  'rasterio', 'geoviews', 'trollimage', 'fsspec', 'bottleneck',
                  'rioxarray', 'pytest', 'pytest-lazy-fixture', 'defusedxml',
                  's3fs', 'eccodes', 'h5netcdf', 'xarray-datatree',
-                 'skyfield', 'ephem', 'pint-xarray', 'astropy']
+                 'skyfield', 'ephem', 'pint-xarray', 'astropy', 'dask-image']
 
 extras_require = {
     # Readers:
@@ -77,6 +77,7 @@
     # Composites/Modifiers:
     'rayleigh': ['pyspectral >= 0.10.1'],
     'angles': ['pyorbital >= 1.3.1'],
+    'filters': ['dask-image'],
     # MultiScene:
     'animations': ['imageio'],
     # Documentation: