A fast method fo applying rigid transformations to volume data

Abstract

Volume rendering is a widespread method for displaying graphical models in fields such as medical visualization or engineering. The required image information is stored in a volume dataset, which is a threedimensional array of voxel intensities. Sometimes it is necessary to transform a given volume dataset by geometric transformations like rotation, scaling or translation. In this process, a new volume dataset containing the transformed volume is generated. Straightforward algorithms with an emphasis on minimizing the resampling error are the standard approach to computing such transformations. In this paper we present a new algorithm, which is significantly faster than these. Since it is based on a simple interpolation scheme, it is useful for generating fast previews. Depending on their size and voxel bit depth, it can even provide realtime transformation of volumes.