def test_matching_order(covariance_transformation: MagicMock,
whitening: MagicMock, center_image: MagicMock,
get_feature_matrix: MagicMock,
feature_dist_matching: FeatureDistMatching) -> None:
mock = Mock()
mock_get_feature_matrix = 'mock_get_feature_matrix'
mock_center_image = 'mock_center_image'
mock_whitening = 'mock_whitening'
mock_covariance_transformation = 'mock_covariance_transformation'
mock.attach_mock(get_feature_matrix, mock_get_feature_matrix)
mock.attach_mock(center_image, mock_center_image)
mock.attach_mock(whitening, mock_whitening)
mock.attach_mock(covariance_transformation, mock_covariance_transformation)
source = TEST_IMAGE_423
reference = TEST_IMAGE_243
feature_dist_matching._matching(source, reference)
# check the call order
expected_calls = [
mock_get_feature_matrix, mock_get_feature_matrix, mock_center_image,
mock_center_image, mock_whitening, mock_covariance_transformation,
mock_covariance_transformation + '().__iadd__',
mock_covariance_transformation + '().__iadd__().reshape'
]
calls = [call[0] for call in mock.mock_calls]
assert calls == expected_calls
def test_whitening_2d() -> None:
test_image = TEST_IMAGE_421
test_image_mat = FeatureDistMatching._get_feature_matrix(test_image)
FeatureDistMatching._center_image(test_image_mat)
result = FeatureDistMatching._whitening(test_image_mat)
np.testing.assert_almost_equal(np.var(result), 1.)
def test_apply_channel_range(
feature_dist_matching: FeatureDistMatching) -> None:
source = TEST_IMAGE_423
reference = TEST_IMAGE_243
source[:, :, 1] *= 255.
source[:, :, 2] *= 254.
source[:, :, 2] -= 127.
reference[:, :, 1] *= 255.
reference[:, :, 2] *= 254.
reference[:, :, 2] -= 127.
feature_dist_matching.channel_ranges = tuple([
ChannelRange(0., 1.),
ChannelRange(0., 255.),
ChannelRange(-127.0, 127.0)
])
result = feature_dist_matching._apply(source, reference)
np.testing.assert_array_equal(source, TEST_IMAGE_423)
np.testing.assert_array_equal(reference, TEST_IMAGE_243)
assert result.shape == source.shape
assert result.dtype == np.float32
assert np.max(result[:, :, 0]) <= 1.
assert 0.1 < np.mean(result[:, :, 0]) < 0.9
assert np.max(result[:, :, 0]) >= 0.
assert np.max(result[:, :, 1]) <= 255.
assert 5. < np.mean(result[:, :, 1]) < 250.
assert np.min(result[:, :, 1]) >= 0.
assert np.max(result[:, :, 2]) <= 127.
assert -120. < np.mean(result[:, :, 2]) < 120.
assert np.min(result[:, :, 2]) >= -127.
def test_whitening() -> None:
test_image = TEST_IMAGE_423
test_image_mat = FeatureDistMatching._get_feature_matrix(test_image)
FeatureDistMatching._center_image(test_image_mat)
result = FeatureDistMatching._whitening(test_image_mat)
np.testing.assert_almost_equal(np.cov(result, rowvar=False),
np.identity(3))
def test_center_image() -> None:
test_image = TEST_IMAGE_423
test_image_mat = FeatureDistMatching._get_feature_matrix(test_image)
test_image_mean = np.mean(test_image_mat, axis=0)
result_image_mat, result_mean = FeatureDistMatching._center_image(
test_image_mat)
np.testing.assert_array_equal(result_mean, test_image_mean)
np.testing.assert_almost_equal(np.mean(result_image_mat, axis=0),
np.zeros(3))
def test_apply_channels(channels: ChannelsType,
feature_dist_matching: FeatureDistMatching) -> None:
source = TEST_IMAGE_423.astype(np.float32)
reference = TEST_IMAGE_243.astype(np.float32)
feature_dist_matching.channels = channels
result = feature_dist_matching._apply(source, reference)
for untouched_channel in {0, 1, 2} - set(channels):
np.testing.assert_array_equal(source[:, :, untouched_channel],
result[:, :, untouched_channel])
def test_matching_2d(channel_range_max: float) -> None:
source = TEST_IMAGE_421 * channel_range_max
reference = TEST_IMAGE_421 * channel_range_max
result = FeatureDistMatching._matching(source, reference)
feature_mat_result = FeatureDistMatching._get_feature_matrix(result)
feature_mat_ref = FeatureDistMatching._get_feature_matrix(reference)
result_mean = np.mean(feature_mat_result, axis=0)
reference_mean = np.mean(feature_mat_ref, axis=0)
np.testing.assert_almost_equal(result_mean, reference_mean)
feature_mat_result -= result_mean
feature_mat_ref -= reference_mean
np.testing.assert_almost_equal(np.var(feature_mat_result),
np.var(feature_mat_ref))
def test_matching(channel_range_max: float) -> None:
source = TEST_IMAGE_423
reference = TEST_IMAGE_243
source[:, :, 1] *= channel_range_max
reference[:, :, 1] *= channel_range_max
result = FeatureDistMatching._matching(source, reference)
feature_mat_result = FeatureDistMatching._get_feature_matrix(result)
feature_mat_ref = FeatureDistMatching._get_feature_matrix(reference)
result_mean = np.mean(feature_mat_result, axis=0)
reference_mean = np.mean(feature_mat_ref, axis=0)
np.testing.assert_almost_equal(result_mean, reference_mean)
feature_mat_result -= result_mean
feature_mat_ref -= reference_mean
np.testing.assert_almost_equal(np.cov(feature_mat_result, rowvar=False),
np.cov(feature_mat_ref, rowvar=False))
def test_apply(feature_dist_matching: FeatureDistMatching, source: np.array,
reference: np.array) -> None:
feature_dist_matching.channels = CHANNELS_DEFAULT[0:source.shape[-1]]
feature_dist_matching.channel_ranges = CHANNEL_RANGES_DEFAULT[0:source.
shape[-1]]
source_copy = np.copy(source)
reference_copy = np.copy(reference)
result = feature_dist_matching._apply(source, reference)
# Check modification of input:
np.testing.assert_array_equal(source, source_copy)
np.testing.assert_array_equal(reference, reference_copy)
# Check result:
assert result.shape == source.shape
assert np.max(result) <= 1.
assert 0.1 < np.mean(result) < 0.9
assert np.min(result) >= 0.
assert result.dtype == np.float32
def test_apply_images(source_path: str, reference_path: str,
feature_dist_matching: FeatureDistMatching) -> \
None:
source = read_image(source_path)
reference = read_image(reference_path)
result = feature_dist_matching._apply(source, reference)
assert result.shape == source.shape
assert np.max(result) <= 1.
assert 0.1 < np.mean(result) < 0.9
assert np.min(result) >= 0.
assert result.dtype == np.float32
with np.testing.assert_raises(AssertionError):
np.testing.assert_array_equal(result, source)
with np.testing.assert_raises(AssertionError):
np.testing.assert_array_equal(result, reference)
def test_covariance_transformation_2d() -> None:
test_image = TEST_IMAGE_421
test_image_mat = FeatureDistMatching._get_feature_matrix(test_image)
FeatureDistMatching._center_image(test_image_mat)
test_image_white = FeatureDistMatching._whitening(test_image_mat)
feature_mat_ref = FeatureDistMatching._get_feature_matrix(
TEST_IMAGE_243[:, :, 0])
result = FeatureDistMatching._covariance_transformation(
test_image_white, feature_mat_ref)
np.testing.assert_almost_equal(np.var(result), np.var(feature_mat_ref))
def test_covariance_transformation() -> None:
test_image = TEST_IMAGE_423
test_image_mat = FeatureDistMatching._get_feature_matrix(test_image)
FeatureDistMatching._center_image(test_image_mat)
test_image_white = FeatureDistMatching._whitening(test_image_mat)
feature_mat_ref = FeatureDistMatching._get_feature_matrix(TEST_IMAGE_243)
result = FeatureDistMatching._covariance_transformation(
test_image_white, feature_mat_ref)
np.testing.assert_almost_equal(np.cov(result, rowvar=False),
np.cov(feature_mat_ref, rowvar=False))
def test_get_feature_matrix(test_image: np.array,
des_shape: Tuple[int, int]) -> None:
result = FeatureDistMatching._get_feature_matrix(test_image)
np.testing.assert_array_equal(result, test_image.reshape(des_shape))