def get_image_quality(
patterns: Union[np.ndarray, da.Array],
frequency_vectors: Optional[np.ndarray] = None,
inertia_max: Union[None, int, float] = None,
normalize: bool = True,
) -> np.ndarray:
"""Compute the image quality in a chunk of EBSD patterns.
The image quality is calculated based on the procedure defined by
Krieger Lassen [Lassen1994]_.
Parameters
----------
patterns
EBSD patterns.
frequency_vectors
Integer 2D array with values corresponding to the weight given
each FFT spectrum frequency component. If None (default), these
are calculated from
:func:`~kikuchipy.util.pattern.fft_frequency_vectors`.
inertia_max
Maximum inertia of the FFT power spectrum of the image. If None
(default), this is calculated from the `frequency_vectors`.
normalize
Whether to normalize patterns to a mean of zero and standard
deviation of 1 before calculating the image quality. Default
is True.
Returns
-------
image_quality_chunk : numpy.ndarray
Image quality of patterns.
"""
dtype_out = np.float64
image_quality_chunk = np.empty(patterns.shape[:-2], dtype=dtype_out)
for nav_idx in np.ndindex(patterns.shape[:-2]):
# Get (normalized) pattern
if normalize:
pattern = pattern_processing.normalize_intensity(
pattern=patterns[nav_idx])
else:
pattern = patterns[nav_idx]
# Compute image quality
image_quality_chunk[nav_idx] = pattern_processing.get_image_quality(
pattern=pattern,
normalize=False,
frequency_vectors=frequency_vectors,
inertia_max=inertia_max,
)
return image_quality_chunk
def test_get_image_quality_pattern(
self, dummy_signal, idx, normalize, frequency_vectors, inertia_max, answer
):
p = dummy_signal.inav[idx].data.astype(np.float32)
iq = get_image_quality(
pattern=p,
normalize=normalize,
frequency_vectors=frequency_vectors,
inertia_max=inertia_max,
)
assert np.allclose(iq, answer, atol=1e-4)
def test_get_image_quality_flat(self):
p = np.ones((1001, 1001)) * 5
iq = get_image_quality(pattern=p, normalize=False)
assert np.allclose(iq, 1, atol=1e-2)
def test_get_image_quality_white_noise(self):
p = np.random.random((1001, 1001))
iq = get_image_quality(pattern=p, normalize=True)
assert np.allclose(iq, 0, atol=1e-2)