Projection and reconstruction-based noise filtering methods in cone beam CT

This thesis has been motivated and developed within a collaboration project between the Department of Mathematics of the University of Bologna and the company Skanray Europe Srl. In detail, this work focuses on Cone Beam CT reconstruction of the human head, with particular interest in detecting small soft tissue elements within white and grey brain matter. This problem is expecially difficult due to the small contrast in the X-ray attenuation coefficient between the two substances to be discriminated. After accurately modelling the data acquisition process (also according to the requirements of the company), we propose a novel reconstruction algorithm based on two subsequent steps. First, the measured and degraded projections are restored, namely the blur and the noise are removed. Second, the restored projections are feeded as input to a variational reconstruction method named TV3D-L2, so as to get an accurate final reconstruciton of the head. Several experimental tests are presented which provide evidence for the proposed approach outperforming classical ones expecially for low-dose CT.