Computed tomography postacquisition data correction for system alignment errors

A critical issue in obtaining high-quality images from a computed tomography system is the mechanical alignment of the data acquisition system. This paper will compare two methods of measuring and correcting errors introduced into second-generation computed tomography data sets from misalignment of the system. Both of these methods use a projection data set from a complete scan of an object as a starting point. One method is an iterative method that looks at residual errors in selected portions of the reconstructed image. The other method requires that the centroids of the projection data coincide with the centroid of the object. It will be shown that the centroid registration algorithm can be used to accurately measure a number of important computed tomography system alignment parameters and is mathematically efficient. The residual error algorithm will be shown to give less satisfactory results.