A framework for nuclear image enhancement based on the Anscomb transform and the bayesian thresholding
Imen Mbazaa, Najla Gribaa, Nawrès Khlifa, Kamel Hamrouni · 2009
Scintigraphic imagery is an important data source since it provides morphological and functional clinical informations. However, scintigraphic images present very bad quality because of several degradation factors. In fact, the recorded data are embedded in noise modelled as the realisation of Poisson process. The aim of this paper is to provide a framework able to enhance nuclear images quality by Poisson intensity estimation. This framework consists, in a first step, of performing variance-stabilizing step for the Poisson process thanks to the Anscombe transformation. So the obtained data can be considered as contaminated by a white Gaussian noise. In a second step, it uses a Bayesian technique inspired of Pizurica approach, known in the literature for exhibiting good results as for white Gaussian noise. In fact, the complex wavelet packets were exploited regarding to Pizurica algorithm.