Computed tomography with view angle estimation using uncertainty quantification

Abstract We consider computed tomography (CT) with uncertain measurement geometry, a focus on the case where view angles are and estimation of these improves reconstruction. propose new reconstruction model corresponding algorithm that has an additional view-angle component, allowing us to determine solely from measured CT data. A key component our approach is we quantify uncertainty via model-discrepancy formulation, take into account in image This generalizes straightforward way other cases geometry. Our method computationally efficient since can utilize block-structure computational problem for both under assumption independent. The joint image/angle non-convex which leads difficulties recently proposed algorithms, demonstrate numerically seems avoid difficulties. Simulations show method, total variation prior reflects phantoms, able achieve reconstructions whose quality similar ones obtained correct (the ideal scenario).