Dual approach for TV-regularized Maximum Likelihood Expectation Maximization in tomography with Poisson data
Résumé
Tomography in nuclear medicine requires resolution of a linear inverse problem. As the radioactive decay follows a Poisson law, the projections of gamma emission from inside the body are also Poisson distributed. In this talk we will introduce a dual algorithm for total variation (TV) Poisson denoising. When combined with MLEM (maximum likelihood expectation maximization) iterations, a fast and convergent algorithm for the estimation of the TV maximum-a-posteriori solution is obtained.