There are several methods available to determine the Centre Of Rotation, COR, and align detectors and X-ray focus to COR in X-ray computed tomography. Some methods use narrow rods/needles or specially made alignment objects or phantoms. In X-ray Micro Computed Tomography (or Computerized Micro Tomography), μCT (CMT), methods using sample projection data for COR measurements are preferred since the replacement of alignment objects with samples often displace translation stages and make the alignment obsolete. To achieve an optimal image quality, precise positioning of COR to the detector and X-ray focus is essential. In μCT this can be accomplished in an alignment procedure using sample projection data prior to scanning. This alignment procedure will add examination time and increase the dose to the sample. Therefore the alignment procedure should incorporate as few projections as possible and be insensitive to noise. Some scanning equipment cannot be modified to implement such alignment procedure but actual COR can be determined from projection data and used in reconstruction. This report introduces a new Translated Opposite Projection, TOP, technique using a pair of opposite parallel projections (180° apart). Two TOP methods are developed: TOPlin using linear interpolation in the spatial domain and TOPfft in the frequency domain. The two TOP methods are compared to two Centre Of Gravity, COG, methods. The two COG methods are: COGsin, an enhancement of a method presented by Hogan et al [Hogan93] and COGopp, a simplification of this method possible with a fixed COR.
In this report all projections in one scan are assumed to have a fixed, COR, as in third (or higher) generation tomography or first (and second generation) tomography if the translation stage errors is negligible. This also means that the rotation stage errors must be negligible. The COGsin is the only method presented here capable of determine a COR for each projection angle, thus allowing for a COR moving as a function of projection angle. The TOP methods normally give better precision with non-ideal projection data compared to the COG methods. Tests using both simulated and scanned projection data indicate that the TOP methods give higher precision in presence of stochastic errors (noise) and system errors like calibration errors. A μCT scan often takes a long time and detector calibration and X-ray intensity profiles might vary with time giving non-stationary system errors during a single scan. If the system errors can be approximated with simple polynomial functions, a new Baseline Restoration, BR, technique can be used together with the TOP methods to reduce COR errors.
Linköping: Linköping University Electronic Press , 1994. , 21 p.