Multilevel algorithm for a Poisson noise removal model with total-variation regularization

Many commonly used models for the fundamental image processing task of noise removal can deal with Gaussian white noise. However such Gaussian models are not effective to restore images with Poisson noise, which is ubiquitous in certain applications. Recently Le-Chartrand-Asaki derived a new data-fitting term in the variational model for Poisson noise. This paper proposes a multilevel algorithm for efficiently solving this variational model. As expected of a multilevel method, it delivers the same numerical solution many orders of magnitude faster than the standard single-level method of coordinate descent time-marching. Supporting numerical experiments on 2D gray scale images are presented.