def get_image_variance(self, dqe):
"""Calculates the variance in scattered intensity as a function of
scattering vector. The calculated variance is normalised by the mean
squared, as is appropriate for the distribution of intensities. This
causes a problem if Poisson noise is significant in the data, resulting
in a divergence of the Poisson noise term. To in turn remove this
effect, we subtract a dqe/mean_dp term (although it is suggested that
dqe=1) from the data, creating a "poisson noise-free" corrected variance
pattern. DQE is fitted to make this pattern flat.
Parameters
----------
dqe : float
Detective quantum efficiency of the detector for Poisson noise
correction.
Returns
-------
varims : ImageVariance
A two dimensional Signal class object containing the mean DP, mean
squared DP, and variance DP, and a Poisson noise-corrected variance
DP.
"""
im = self.signal.T
mean_im = im.mean((0, 1))
meansq_im = Signal2D(np.square(im.data)).mean((0, 1))
normvar = (meansq_im.data / np.square(mean_im.data)) - 1.0
var_im = Signal2D(normvar)
corr_var_array = normvar - (np.divide(dqe, mean_im.data))
corr_var_array[np.invert(np.isfinite(corr_var_array))] = 0
corr_var = Signal2D(corr_var_array)
varims = stack((mean_im, meansq_im, var_im, corr_var))
sig_x = varims.data.shape[1]
sig_y = varims.data.shape[2]
iv = ImageVariance(varims.data.reshape((2, 2, sig_x, sig_y)))
iv = transfer_navigation_axes_to_signal_axes(iv, self.signal)
return iv
def test_get_image_variance_signal(self, diffraction_pattern):
imvar = ImageVariance(diffraction_pattern)
assert isinstance(imvar, ImageVariance)