This article shows that re-normalizing the interpolation kernel for a const
ant integral can make a significant improvement in performance of sine inte
rpolation methods. A comparison was performed between standard and re-norma
lized sine kernels of various sizes using data from four commonly used magn
etic resonance (MR) imaging sequences. Standard rotations were performed an
d compared with a "gold standard" data set generated by use of a large (13
x 13 x 13) sine kernel. Measurements of systematic pixel intensity offset e
rror and variance of generated residuals were used to estimate resultant in
terpolation error, Theoretical estimates of the consequent savings in compu
tation time were compared with the measured time required for each algorith
m and with the automated image registration (AIR) program. The use of a sma
ll (5 x 5 x 5) re-normalized kernel produced relative errors comparable to
those in the gold standard data set, allowing saving in computation time of
up to 30 times in comparison with standard sine interpolation. This approa
ch brings the implementation of MR volume re-slicing much closer to the dem
ands of a clinical environment. (C) 1999 Wiley-Liss, Inc.