Errors in projection position arise when the resolution required by to
mographic systems approaches the tolerance of the translational stage.
These shifts in the projection data, commonly known as backlash, appe
ar as jagged edges in a sinogram. If these random shifts are not remov
ed, the resulting reconstruction is blurred. This reduces the effectiv
e resolution of the system preventing viewing of single pixel events w
ithin the sample. An iterative method is presented for removing the tr
anslational backlash from projection data. The proposed method finds t
he center of gravity for all projections within a fixed window of the
projection. The center of gravity of the projection will follow a sinu
soidal path through the sinogram. A minimum mean square error (MMSE) f
it to a sinusoid is made. If the center of gravity of a projection doe
s not match the fit value, the entire projection is shifted into the c
orrect position using linear interpolation. Since the shift of a proje
ction introduces new data (that includes random noise) into the fixed
calculation window, the center of gravity is again calculated, and the
appropriate shift is made. This process is repeated until an acceptab
le error is reached. Since each shift uses linear interpolation, the d
ata is blurred with each iteration. In order to minimize the amount of
blurring, the shift values for each iteration are saved. The saved va
lues are added and single shifts of the original projection data are m
ade. An example of translational stage error removal is included.