GammaNoise#

class deepinv.physics.GammaNoise(l=1.0)[source]#

Bases: NoiseModel

Gamma noise \(y = \mathcal{G}(\ell, x/\ell)\)

Follows the shape, scale parameterization of the Gamma distribution, where the mean is given by \(x\) and the variance is given by \(x/\ell\), see https://en.wikipedia.org/wiki/Gamma_distribution for more details.

Distribution for modelling speckle noise (eg. SAR images), where \(\ell>0\) controls the noise level (smaller values correspond to higher noise).

Warning

This noise model does not support the random number generator.

Parameters:: l (float, torch.Tensor) – noise level.

forward(x, l=None, **kwargs)[source]#

Adds the noise to measurements x

Parameters:

x (torch.Tensor) – measurements
l (None, float, torch.Tensor) – noise level. If not None, it will overwrite the current noise level.

Returns:

noisy measurements

update_parameters(l=None, **kwargs)[source]#

Updates the noise level.

Parameters:: ell (float, torch.Tensor) – noise level.