Methodology for the evaluation of CT image quality in dimensional metrology

In the last years, x-ray computed tomography (CT) has gained more and more importance in metrology. However, computed tomography is a complex and indirect measurement procedure. Many factors contribute to the measurement result, which makes it difficult for the user to relate cause and effect. For example, the userset scan parameters significantly influence the measurement result. However, the selection of scan parameters is still based on the experience of the CT user, while the influence of the parameter choice on the measurement result cannot be quantified. This leads to diverging and non-optimal scan results. The quality of the CT scan can only be evaluated afterwards, e.g. by visual inspection of the reconstructed volume. Taking into account that the process chain of CT measurements is highly complex, the very first step is the generation of reliable high quality projections that can then be fed into the reconstruction. The quality of these projections can be described quantitatively by image quality measures. Thus, in this paper, an approach regarding projection based evaluation of CT image quality in micro computed tomography is presented. By performing a set of experiments, the influence of projection image quality on the reconstructed volume and respectively on the measurement result is evaluated. A derived model relates the projection quality measures to the actual measurement error of the CT scan. Hence, the proposed approach defines a reliable, combined quality measure, which is based on a small number of projections acquired before the actual scan. An algorithm evaluates the quality of those projection for every workpiece that has to be measured. Based on the evaluation, it varies the scan parameters, until an optimal projection quality is reached and a reduced measurement error in the CT scan is achieved.