RandomPhaseError

class deepinv.transform.RandomPhaseError(*args, scale=0.2, **kwargs)

Bases: Transform

Random phase error transform.

This transform is specific to MRI problems, and adds a phase error to k-space using:

\(Ty=\exp(-i\phi_k)y\) where \(\phi_k=\pilpha s_e\) if \(k\) is an even index, or \(\phi_k=\pilpha s_o\) if odd, and where \(lpha\) is a scale parameter, and \(s_o,s_e\sim U(-1,1)\).

This transform is reproducible: for given param dict se, so, the transform is deterministic.

Parameters:

scale (int, tuple[int, int]) – scale parameters \(s_e\) and \(s_o\) or range to pick randomly.