lentil.detector.shot_noise#

lentil.detector.shot_noise(img, method='poisson', seed=None)[source]#

Apply shot noise to an image

Parameters:

img (array_like) – Array of counts. All values must be >= 0.
method ('poisson' or 'gaussian') – Noise method.
seed (None, int, or array_like, optional) – Random seed used to initialize the pseudo-random number generator. If seed is None (default), the seed will be randomly generated from /dev/urandom if available or the system clock.

Returns:

img – Array of noisy counts

Return type:

ndarray

Notes

The output of the Poisson distribution is limited to the range of the C int64 type. A ValueError is raised when img contains values within 10 sigma of the maximum representable value (lam > 9.223372006484771e+18).

For sufficiently large values of \(\lambda\), (say \(\lambda > 1000\)), the Normal(\(\mu = \lambda\), \(\sigma^2 = \lambda\)) distribution is an excellent approximation to the Poisson(\(\lambda\)) distribution and is about 25% faster to compute.