Since its invention in the late 1960s, computed tomography is nowadays an indispensable imaging technology in medicine, non-destructive testing and materials research. A challenge often encountered in imaging techniques is the need to quantify the changes between two states of the object under inspection. This includes e.g. the deformation induced to a sample by external forces or the misalignment of a specimen between two consecutive CT scans. In conventional approaches, this deformation estimation is done using reconstructed image data. In this thesis, new approaches are developed which allow us to perform the deformation estimation directly from the tomographic projections, avoiding the computationally expensive reconstruction process.
In a first approach, we investigate the eect of globally ane transformations to tomographic
projections. We show that any ane transformation can be analytically compensated in 2d/3d parallel beam and fan beam projection geometries. A multiscale optimization framework is presented for the estimation of the ane deformation parameters. We use simulated projection data of several test images to assess the performance of our algorithm in terms of estimation accuracy, inuence of noise and the number of projection angles used. Also, we discuss the computational complexity and we formulate a necessary (but not sucient) property which the assumed deformation model must fulll. The investigation is concluded by a discussion of the cone beam geometry and the presentation of an application example, namely the estimation of motion in the presence of streak artifacts caused by high density regions in the sample.