EM algorithm for estimating poisson measurement noise
Garry Einicke · 2014
In the standard minimum-variance filter recursions it is routinely assumed that the noises are zero-mean and white. In image restoration applications, the data can be contaminated with (nonzero-mean) Poisson noise. This paper introduces the minimum-variance filter for the case where the measurement noise includes a Poisson-distributed component. An EM algorithm for estimating the Poisson noise intensity is described. Conditions for the convergence of the algorithms are also investigated. An image restoration example is presented which demonstrates the efficacy of the described method.