Non-ideal projection data in X-ray computed tomography

This thesis deals with various aspects of X-ray computed tomography (CT) from non-ideal projection data that do not permit the successful application of standard image reconstruction methods. A versatile calibration method was implemented allowing an adequate correction for geometric flex effects introduced by non-ideal geometry during data acquisition. A filtered backprojection algorithm was modified such that the geometric information from the calibration can be considered with high accuracy and in a computationally efficient way. The methods proposed here were evaluated by means of computer simulations and measurements using two experimental cone-beam CT scanners. Results are presented that demonstrate their excellent performance. Furthermore, a new retrospective image restoration scheme was developed in order to tackle the problem of inconsistent projection data due to internal organ motion. This scheme was applied to simulated projections as well as clinical CT data of a beating heart. The results of these investigations indicate that the image restoration approach is a promising alternative to existing, well-established image reconstruction methods if considerable organ motion occurs during data acquisition