Merge pull request #208 from DiamondLightSource/flatsdarks_scaler

normalisation method parameters for scaling flats/darks

Author: dkazanc

httomolibgpu/prep/normalize.py

@@ -43,6 +43,8 @@ def normalize(
     data: cp.ndarray,
     flats: cp.ndarray,
     darks: cp.ndarray,
+    flats_multiplier: float = 1.0,
+    darks_multiplier: float = 1.0,
     cutoff: float = 10.0,
     minus_log: bool = True,
     nonnegativity: bool = False,
@@ -60,13 +62,17 @@ def normalize(
         3D flat field data as a CuPy array.
     darks : cp.ndarray
         3D dark field data as a CuPy array.
-    cutoff : float, optional
+    flats_multiplier: float
+        A multiplier to apply to flats, can work as an intensity compensation constant.
+    darks_multiplier: float
+        A multiplier to apply to darks, can work as an intensity compensation constant.
+    cutoff : float
         Permitted maximum value for the normalised data.
-    minus_log : bool, optional
+    minus_log : bool
         Apply negative log to the normalised data.
-    nonnegativity : bool, optional
+    nonnegativity : bool
         Remove negative values in the normalised data.
-    remove_nans : bool, optional
+    remove_nans : bool
         Remove NaN and Inf values in the normalised data.
 
     Returns
@@ -82,6 +88,9 @@ def normalize(
     mean(darks, axis=0, dtype=float32, out=dark0)
     mean(flats, axis=0, dtype=float32, out=flat0)
 
+    dark0 *= darks_multiplier
+    flat0 *= flats_multiplier
+
     kernel_name = "normalisation"
     kernel = r"""
         float denom = float(flats) - float(darks);

tests/test_prep/test_normalize.py

@@ -35,6 +35,26 @@ def test_normalize(data, flats, darks, ensure_clean_memory):
     assert data_normalize.flags.c_contiguous
 
 
+def test_normalize_scale(data, flats, darks, ensure_clean_memory):
+    # --- testing normalize  ---#
+    data_normalize = normalize(
+        cp.copy(data),
+        flats,
+        darks,
+        flats_multiplier=10.0,
+        darks_multiplier=10.0,
+        minus_log=True,
+    ).get()
+
+    assert data_normalize.dtype == np.float32
+
+    assert_allclose(np.mean(data_normalize), 2.5918312, rtol=1e-06)
+    assert_allclose(np.mean(data_normalize, axis=(1, 2)).sum(), 466.5298, rtol=1e-06)
+    assert_allclose(np.median(data_normalize), 2.3198225, rtol=1e-06)
+    assert_allclose(np.std(data_normalize), 0.5243825, rtol=1e-06)
+    assert data_normalize.flags.c_contiguous
+
+
 @pytest.mark.perf
 def test_normalize_performance(ensure_clean_memory):
     # Note: low/high and size values taken from sample2_medium.yaml real run