Tomography: the Cormack algorithm revisited

Summary form only given. The Cormack algorithm for doing tomography is easily extended from parallel geometry to fan geometry. After a Fourier transform of the data (in rotation angle space) one is required to solve a truncated system before back Fourier transforming to object space. This is carried out by doing a Galerkin discretization and converting the system to a system of matrix equations, which are solved by singular value decomposition (SVD). All of the SVD work on all N+1 kernels, can be done once and stored. Some reconstructions derived from very noisy data have been obtained.<
>