def test_add_noise_filter_snr_zero(image_container):
""" Checks that the output image is equal to the input image when snr=0 """
# calculate using the filter
add_noise_filter = AddNoiseFilter()
add_noise_filter.add_input("image", image_container)
add_noise_filter.add_input("snr", 0.0)
np.random.seed(RANDOM_SEED) # set seed so RNG is in the same state
add_noise_filter.run()
# image_with_noise and image_with_noise_container.image should be equal
numpy.testing.assert_array_equal(image_container.image,
add_noise_filter.outputs["image"].image)
def _create_filter_block(self):
"""Fourier transforms the input and reference images, calculates
the noise amplitude, adds this to the FT of the input image, then
inverse fourier transforms to obtain the output image"""
input_image: BaseImageContainer = self.inputs[self.KEY_IMAGE]
# if self.inputs["snr"] == 0 then the input image should just be
# passed through to the output.
# Because this is a filter block it needs to be done by another filter - fortunately
# the AddNoiseFilter does just this when AddNoiseFilter.inputs["snr"] = 0
if self.inputs[self.KEY_SNR] == 0:
add_noise_filter = AddNoiseFilter()
add_noise_filter.add_input(self.KEY_IMAGE, input_image)
add_noise_filter.add_input(self.KEY_SNR, self.inputs[self.KEY_SNR])
return add_noise_filter
# snr is greater than 0 so run the filter block normally
# Fourier transform the input image
image_fft_filter = FftFilter()
image_fft_filter.add_input(self.KEY_IMAGE, input_image)
# Create the noise filter
add_noise_filter = AddNoiseFilter()
add_noise_filter.add_parent_filter(image_fft_filter)
add_noise_filter.add_input(self.KEY_SNR, self.inputs[self.KEY_SNR])
# If present load the reference image, if not, copy the input_image
if self.KEY_REF_IMAGE in self.inputs:
add_noise_filter.add_input(
self.KEY_REF_IMAGE, self.inputs[self.KEY_REF_IMAGE]
)
else:
add_noise_filter.add_input(self.KEY_REF_IMAGE, self.inputs[self.KEY_IMAGE])
# Inverse Fourier Transform and set the output
ifft_filter = IfftFilter()
ifft_filter.add_parent_filter(add_noise_filter)
return ifft_filter
def test_add_noise_filter_with_mock_data_complex_image(
complex_image_container):
""" Test the add noise filter with a complex image """
np.random.seed(RANDOM_SEED)
# calculate manually
image_with_noise = add_noise_function(
complex_image_container.image,
SNR_VALUE,
)
image_with_noise_2 = add_noise_function(
complex_image_container.image,
SNR_VALUE,
)
print(
f"manual snr = {calculate_snr_function(image_with_noise,image_with_noise_2)}"
)
# calculate using the filter
add_noise_filter = AddNoiseFilter()
add_noise_filter.add_input("image", complex_image_container)
add_noise_filter.add_input("snr", SNR_VALUE)
np.random.seed(RANDOM_SEED) # set seed so RNG is in the same state
add_noise_filter.run()
image_with_noise_container = add_noise_filter.outputs["image"].clone()
# image_with_noise and image_with_noise_container.image should be equal
numpy.testing.assert_array_equal(image_with_noise,
image_with_noise_container.image)
# Run again and then check the SNR
add_noise_filter = AddNoiseFilter()
add_noise_filter.add_input("image", complex_image_container)
add_noise_filter.add_input("snr", SNR_VALUE)
add_noise_filter.run()
measured_snr = calculate_snr_function(
image_with_noise_container.image,
add_noise_filter.outputs["image"].image,
)
print(f"calculated snr = {measured_snr}, desired snr = {SNR_VALUE}")
# This should be almost equal to the desired snr
numpy.testing.assert_array_almost_equal(measured_snr, SNR_VALUE, 0)
def test_add_noise_filter_with_mock_data_complex_image_inverse_reference_spatial(
complex_image_container, ft_complex_image_container):
"""Test the add noise filter with a complex image INVERSE_DOMAIN,
complex reference in the SPATIAL_DOMAIN
"""
np.random.seed(RANDOM_SEED)
# calculate manually
image_with_noise = add_noise_function(
ft_complex_image_container.image,
SNR_VALUE,
complex_image_container.image,
np.sqrt(complex_image_container.image.size),
)
image_with_noise_2 = add_noise_function(
ft_complex_image_container.image,
SNR_VALUE,
complex_image_container.image,
np.sqrt(complex_image_container.image.size),
)
print(
f"manual snr = {calculate_snr_function(image_with_noise,image_with_noise_2)}"
)
# calculate using the filter
add_noise_filter = AddNoiseFilter()
add_noise_filter.add_input("image", ft_complex_image_container)
add_noise_filter.add_input("snr", SNR_VALUE)
add_noise_filter.add_input("reference_image", complex_image_container)
np.random.seed(RANDOM_SEED) # set seed so RNG is in the same state
add_noise_filter.run()
image_with_noise_container = add_noise_filter.outputs["image"].clone()
# image_with_noise and image_with_noise_container.image should be equal
numpy.testing.assert_array_equal(image_with_noise,
image_with_noise_container.image)
# Run again and then check the SNR
add_noise_filter = AddNoiseFilter()
add_noise_filter.add_input("image", ft_complex_image_container)
add_noise_filter.add_input("snr", SNR_VALUE)
add_noise_filter.add_input("reference_image", complex_image_container)
add_noise_filter.run()
measured_snr = calculate_snr_function(
image_with_noise_container.image,
add_noise_filter.outputs["image"].image,
)
print(f"calculated snr = {measured_snr}, desired snr = {SNR_VALUE}")
# This isn't equal to the desired SNR
with numpy.testing.assert_raises(AssertionError):
numpy.testing.assert_array_almost_equal(measured_snr, SNR_VALUE, 0)
def test_add_noise_filter_validate_inputs():
"""Check a FilterInputValidationError is raised when the inputs
to the add commplex noise filter are incorrect or missing"""
noise_filter = AddNoiseFilter()
noise_filter.add_input("snr", 1)
with pytest.raises(FilterInputValidationError):
noise_filter.run() # image not defined
noise_filter.add_input("image", 1)
with pytest.raises(FilterInputValidationError):
noise_filter.run() # image wrong type
noise_filter = AddNoiseFilter()
noise_filter.add_input("image",
NumpyImageContainer(image=np.zeros((32, 32, 32))))
with pytest.raises(FilterInputValidationError):
noise_filter.run() # snr not defined
noise_filter.add_input("snr", "str")
with pytest.raises(FilterInputValidationError):
noise_filter.run() # snr wrong type
noise_filter = AddNoiseFilter()
noise_filter.add_input("image",
NumpyImageContainer(image=np.zeros((32, 32, 32))))
noise_filter.add_input("snr", 1)
noise_filter.add_input("reference_image", 1)
with pytest.raises(FilterInputValidationError):
noise_filter.run() # reference_image wrong type
noise_filter = AddNoiseFilter()
noise_filter.add_input("image",
NumpyImageContainer(image=np.zeros((32, 32, 32))))
noise_filter.add_input("snr", 1)
noise_filter.add_input("reference_image",
NumpyImageContainer(image=np.zeros((32, 32, 31))))
with pytest.raises(FilterInputValidationError):
noise_filter.run() # reference_image wrong shape