Merge pull request #49 from WEHI-ResearchComputing/41-allow-16-bit-mrc-input-for-denoising

41 allow 16 bit mrc input for denoising

Author: MihinP

partinet/process_utils/guided_denoiser.py

@@ -20,6 +20,9 @@ def transform(image: np.ndarray) -> np.ndarray:
     """
     i_min = image.min()
     i_max = image.max()
+    if i_max == i_min:
+        # avoid division by zero; return a zero array when input is constant
+        return np.zeros_like(image, dtype=np.uint8)
     image = ((image - i_min) / (i_max - i_min)) * 255
     return image.astype(np.uint8)
 
@@ -28,12 +31,20 @@ def standard_scaler(image: np.ndarray) -> np.ndarray:
     """
     Apply Gaussian blur and standardize the image to have zero mean and unit variance.
 
+    The input is cast to ``float32`` before any OpenCV operations to avoid the
+    ``CV_16F`` kernel-type error that occurs when processing 16‑bit micrographs
+    (see issue #41). After blurring and normalization we transform the result to
+    eight‑bit for downstream filters.
+
     Args:
         image (np.ndarray): Input image array.
 
     Returns:
         np.ndarray: Scaled and transformed image.
     """
+    # convert to a supported floating point type for OpenCV kernels
+    image = image.astype(np.float32)
+
     kernel_size = 9
     image = cv2.GaussianBlur(image, (kernel_size, kernel_size), 0)
     mu = np.mean(image)
@@ -150,6 +161,12 @@ def denoise(image_path: str) -> np.ndarray:
     """
     kernel = gaussian_kernel(kernel_size=9)
     image = mrcfile.read(image_path)
+
+    # some MRCs are stored as 16‑bit integers; ensure we work in float32 so that
+    # subsequent OpenCV calls (GaussianBlur, etc.) don't raise the ktype error
+    # described in https://github.com/WEHI-ResearchComputing/PartiNet/issues/41
+    image = image.astype(np.float32)
+
     image = image.T
     image = np.rot90(image)
     normalized_image = standard_scaler(np.array(image))

pyproject.toml

@@ -40,6 +40,7 @@ dependencies = [
   "tensorboard",
   "scikit-learn",
   "mrcfile",
+  "pytest",
 ]
 
 [project.scripts]

tests/test_guided_denoiser.py

@@ -0,0 +1,51 @@
+import os
+import tempfile
+
+import numpy as np
+import mrcfile
+import pytest
+
+from partinet.process_utils.guided_denoiser import denoise, transform, standard_scaler
+
+
+def _write_temp_mrc(array: np.ndarray, dtype: np.dtype) -> str:
+    """Write ``array`` to a temporary .mrc file and return its path."""
+    fd, path = tempfile.mkstemp(suffix=".mrc")
+    os.close(fd)
+    with mrcfile.new(path, overwrite=True) as mrc:
+        mrc.set_data(array.astype(dtype))
+    return path
+
+
+@ pytest.mark.parametrize("dtype", [np.uint16, np.int16, np.float32])
+def test_denoise_handles_various_dtypes(dtype):
+    """The ``denoise`` pipeline should accept 16-bit and 32-bit inputs without
+    throwing OpenCV kernel-type errors (issue #41).
+    """
+    data = (np.random.rand(32, 32) * 255).astype(dtype)
+    path = _write_temp_mrc(data, dtype)
+
+    try:
+        out = denoise(path)
+        assert isinstance(out, np.ndarray)
+        # our pipeline always returns 8-bit data
+        assert out.dtype == np.uint8
+        assert out.shape == data.shape
+    finally:
+        os.unlink(path)
+
+
+def test_transform_stable_when_constant():
+    """``transform`` should not divide by zero if image has no contrast."""
+    arr = np.full((4, 4), 100, dtype=np.float32)
+    out = transform(arr)
+    assert out.dtype == np.uint8
+    assert np.all(out == 0)
+
+
+def test_standard_scaler_normalises():
+    arr = np.arange(25, dtype=np.float32).reshape(5, 5)
+    scaled = standard_scaler(arr)
+    assert scaled.dtype == np.uint8
+    # scaled values should not all be equal
+    assert scaled.std() > 0