Investigating tomographic reconstruction with a priori geometrical information

A novel strategy to perform tomographic image reconstruction is presented, based on the integration of a priori information about the target image. Such information may come from a different imaging tool or a synthetic model. For a given image quality, providing a priori image information reduces the amount of image information to be reconstructed. According to the data processing inequality this requires less input data or physical measurements, therefore reducing exposure to ionising radiation. A prototype algorithm is described, consisting of a penalized ART where some a priori edge information is encoded in an inhomogeneous, anisotropic smoothing kernel. The algorithm is tested on a 2-dimensional set-up based on the Shepp-Logan phantom.