def test_1_prob(self, im_shape, im_type, channel_wise):
im = self.get_data(im_shape, im_type)
intensity_range = [14, 14]
t = RandKSpaceSpikeNoise(1.0, intensity_range, channel_wise)
out = t(im)
base_t = KSpaceSpikeNoise(t.sampled_locs, [14])
out = out - base_t(im)
assert_allclose(out, im * 0, type_test="tensor")
def test_1_prob(self, im_shape, as_tensor_output, as_tensor_input,
channel_wise):
im = self.get_data(im_shape, as_tensor_input)
intensity_range = [14, 14]
t = RandKSpaceSpikeNoise(1.0, intensity_range, channel_wise,
as_tensor_output)
out = t(im)
base_t = KSpaceSpikeNoise(t.sampled_locs, [14], as_tensor_output)
out = out - base_t(im)
np.testing.assert_allclose(out, im * 0)
def test_1_prob(self, im_shape, im_type, channel_wise):
im = self.get_data(im_shape, im_type)
intensity_range = [14, 14]
t = RandKSpaceSpikeNoise(1.0, intensity_range, channel_wise)
out = t(im)
base_t = KSpaceSpikeNoise(t.sampled_locs, [14])
out = out - base_t(im)
self.assertEqual(type(im), type(out))
if isinstance(out, torch.Tensor):
self.assertEqual(out.device, im.device)
im, out = im.cpu(), out.cpu()
np.testing.assert_allclose(out, im * 0)
def test_same_result(self, im_shape, as_tensor_output, as_tensor_input):
im = self.get_data(im_shape, as_tensor_input)
loc = [0, int(im.shape[1] / 2), 0
] if len(im_shape) == 2 else [0, int(im.shape[1] / 2), 0, 0]
k_intensity = 10
t = KSpaceSpikeNoise(loc, k_intensity, as_tensor_output)
out1 = t(deepcopy(im))
out2 = t(deepcopy(im))
np.testing.assert_allclose(out1, out2)
self.assertIsInstance(out1,
torch.Tensor if as_tensor_output else np.ndarray)
def test_highlighted_kspace_pixel(self, im_shape, as_tensor_output,
as_tensor_input):
im = self.get_data(im_shape, as_tensor_input)
loc = [0, int(im.shape[1] / 2), 0
] if len(im_shape) == 2 else [0, int(im.shape[1] / 2), 0, 0]
k_intensity = 10
t = KSpaceSpikeNoise(loc, k_intensity, as_tensor_output)
out = t(im)
n_dims = len(im_shape)
out_k = fftshift(fftn(out, axes=tuple(range(-n_dims, 0))),
axes=tuple(range(-n_dims, 0)))
log_mag = np.log(np.absolute(out_k))
np.testing.assert_allclose(k_intensity, log_mag[tuple(loc)], 1e-4)
def test_same_result(self, im_shape, im_type, k_intensity):
im = self.get_data(im_shape, im_type)
loc = [0, int(im.shape[1] / 2), 0
] if len(im_shape) == 2 else [0, int(im.shape[1] / 2), 0, 0]
t = KSpaceSpikeNoise(loc, k_intensity)
out1 = t(deepcopy(im))
out2 = t(deepcopy(im))
if isinstance(out1, torch.Tensor):
out1 = out1.cpu()
out2 = out2.cpu()
np.testing.assert_allclose(out1, out2)
def test_highlighted_kspace_pixel(self, im_shape, as_tensor_input,
k_intensity):
im = self.get_data(im_shape, as_tensor_input)
loc = [0, int(im.shape[1] / 2), 0
] if len(im_shape) == 2 else [0, int(im.shape[1] / 2), 0, 0]
t = KSpaceSpikeNoise(loc, k_intensity)
out = t(im)
if isinstance(out, torch.Tensor):
out = out.cpu()
if k_intensity is not None:
n_dims = len(im_shape)
out_k = fftshift(fftn(out, axes=tuple(range(-n_dims, 0))),
axes=tuple(range(-n_dims, 0)))
log_mag = np.log(np.absolute(out_k))
np.testing.assert_allclose(k_intensity, log_mag[tuple(loc)], 1e-4)