Matt,
This can be done in two steps, i.e., first flip the images and then correct for reorientation. To do that, you can use TensorVolumeToImage to split a tensor volume into its components. Next, you flip each component images, which I imagine can be done with FSL. After that, you can piece the images back together with ImageToTensorVolume. Finally, to correct for reorientation, you can use TVReorient, which applies an appropriate rotation, specified via Euler angles, to each tensor. TVReorient is described in the following tutorial.
http://groups.google.com/group/dtitk/web/visualization-of-dti-volumesI'm not sure what you try to do exactly. Often, you can save yourself the hassle either by doing all these flipping business before you do the tensor reconstruction or by waiting till all the analyses are done and only flipping the results for visualization.