LogPoissonLikelihood

class deepinv.optim.LogPoissonLikelihood(N0=1024.0, mu=0.02)

Bases: DataFidelity

Log-Poisson negative log-likelihood.

\[\datafid{z}{y} = N_0 (1^{\top} \exp(-\mu z)+ \mu \exp(-\mu y)^{\top}x)\]

Corresponds to LogPoissonNoise with the same arguments N0 and mu. There is no closed-form of prox_d known.

Parameters:

• N0 (float) – average number of photons

• mu (float) – normalization constant

d(x, y)

Computes the data fidelity distance \(\distance{u}{y}\).

Parameters:

• u (torch.Tensor) – Variable \(u\) at which the distance function is computed.

• y (torch.Tensor) – Data \(y\).

Returns:

(torch.Tensor) data fidelity \(\distance{u}{y}\).

Examples using LogPoissonLikelihood:

Patch priors for limited-angle computed tomography

Patch priors for limited-angle computed tomography