Sparsity based Poisson denoising
Raja Giryes, Michael Elad · 2012
Sparsity based techniques have been widely used for image denoising. In this work we focus on Poisson noise and propose initial stages for a new strategy for its removal. We start with a method that removes the noise by converting it into an additive Gaussian noise using the Anscombe transform, applying a variant of the OMP-denoising algorithm. Then, following the recent work by Salmon et. al., we bypass the need for the Anscombe transform and rely directly on the noise statistics. The new strategy is shown to lead to near state-of-the-art results.