You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When using TransformJ rotate on a stack which has non-cubical voxels (in my case the distance between planes in z is more than 5x the pixel length in x or y) the plugin assumes that voxels are cubical, so x, y, and z size are the same.
This can be reproduced using for example the 'first-instar-brain.tif' sample shipped with Fiji. When I run TransformJ rotate on this file and set a rotation around y of 90 degrees, the voxel size remains unchanged and Fiji reports a width of 29.96 microns for the images in the stack. Since the original stack is composed of 56 planes with 2 micron inbetween them, this can not be true. TransformJ Turn does it correctly, and after turning the stack 90 degrees in y Fiji reports a width of 112 microns, and also a correct voxel size.
This is not a bug but a "feature" (although perhaps an inconvenient one). The TransformJ Rotate plugin simply assumes that voxels are isotropic. If they are not, one can use the TransformJ Affine plugin, which also allows scaling at the same time, to get a correctly rotated volume with isotropic voxels. In a future release I will try to deal with nonisotropic voxels in the TransformJ Rotate plugin.
Reported by Laurent Holtzer:
This is a follow-up on my post to the imagej mailing list, since I believe it is a real bug in Fiji.
When using TransformJ rotate on a stack which has non-cubical voxels (in my case the distance between planes in z is more than 5x the pixel length in x or y) the plugin assumes that voxels are cubical, so x, y, and z size are the same.
This can be reproduced using for example the 'first-instar-brain.tif' sample shipped with Fiji. When I run TransformJ rotate on this file and set a rotation around y of 90 degrees, the voxel size remains unchanged and Fiji reports a width of 29.96 microns for the images in the stack. Since the original stack is composed of 56 planes with 2 micron inbetween them, this can not be true. TransformJ Turn does it correctly, and after turning the stack 90 degrees in y Fiji reports a width of 112 microns, and also a correct voxel size.
Migrated-From: http://fiji.sc/bugzilla/show_bug.cgi?id=674
The text was updated successfully, but these errors were encountered: