where the real changes are. The TransformJ plugin classes are basically
just interfaces to the imagescience algorithms.
> Dear Erik,
>
> thank you for the update, it looks promising!
> Did you also make changes in imagescience itself concerning anisotropic
> image handling? Your huge commit on github [1] mixes cosmetic changes
> with meaningful code changes, so it is a little hard to grasp what
> actually changed.
> It would be great if you could shortly comment on that.
>
> Regards,
> Jan
>
> [1]:
>
https://github.com/fiji/imagescience/commit/46f778a29d7097d9b6f785484a5be7dc048a4718>
>
> On 22.08.2015 11:53, Erik Meijering wrote:
>> Dear TransformJ users,
>>
>> TransformJ 4.0.0 is now available
>> (
http://www.imagescience.org/meijering/software/transformj/) for ImageJ
>> users and will soon also appear in the Fiji distribution.
>>
>> It is a long-awaited major update. So long, in fact, that in the
>> meantime many of you probably got used to the shortcomings of the
>> package and found ways to work around them. Needless to say these
>> work-arounds will no longer work with the new version, so you may need
>> to update your code/macros a bit.
>>
>> Apart from some cosmetic changes the most important fundamental change
>> is that the Affine and Rotate plugins are now able to properly work with
>> image stacks having anisotropic voxels. This means you no longer need to
>> first resample your data to isotropic voxels (if this was part of your
>> pipeline you will need to remove that step). The plugins offer the
>> possibility to either sample the transformed image on the same grid
>> (same anisotropy) as the input image (this is the default mode) or to
>> resample the stack isotropically while transforming (an option to be
>> selected by the user).
>>
>> Even though this means TransformJ now finally works the way it should
>> have from the beginning, perhaps some of you would like to keep the old
>> behavior of the plugins. This is possible. The only thing you need to do
>> then is temporarily set the voxel width/height/depth to 1.0 (which is
>> what TransformJ has been assuming all along previously), then perform
>> the transformation, and afterwards put these properties back to their
>> old values.
>>
>> For consistency with the Affine plugin (the most generic plugin of the
>> package), two other changes were introduced. First, the Scale plugin by
>> default no longer rescales the voxel properties in order to retain the
>> physical image dimensions, but this is now an option (to be selected by
>> the user). Second, the Translate plugin now assumes the translation
>> distances to be in physical units (determined by the voxel properties),
>> but an option has been added (to be selected by the user) to take these
>> distances as voxel units.
>>
>> Have fun transforming your data!
>>
>> Erik
>>
>
>