Julian Rasch

Joint Reconstruction via Coupled Bregman Distances

jeudi 26 janvier 2017, 13h30 - 14h30

Salle du conseil, espace Turing

Joint image reconstruction methods exploit similar geometric struc- tures to provide reasonable reconstructions in presence of degraded data. We use the infimal convolution of Bregman distances with respect to the total variation to couple edge information during a joint reconstruction. In particular, unlike most existing methods relying on gradient informa- tion, it does not require a scaling of gradients to similar size. This makes it especially suitable for imaging problems with different contrasts on dif- ferent scales. As a particular application, we show results on artificial PET-MR data.