PPT Slide

In the context of Bayessian regularization, the assumption

behind the quadratic Data term, it is that the error:

as a subproblem

has a Gaussian (normal) distribution.

By cause of discontinuities in the phase and noise ,

err is “corrupted” by outliers: