Update Correct_3D_drift.py

[tischi]

Encouraged by Albert Cardona I added functionality to Correct_3D_drift.py. This was motivated by various fluorescence microscopy data requiring these additions (the changed code has same macro recording behavior as old code => if Fiji users have macros using Correct_3D_drift they will not notice the changes).

Changed code:

• also works now for 2D time-series (used to be 3D only)
• added option: measure drift on multiple timescales. This allows to also find slow drift components of less than 1 pixel per frame (used to not be detected).
• added option: correct sub-pixel drift computing the shifted images using the Translate() function of TransformJ (used to do integer shifts only)
• added option: compute drift on edge filtered images (for some of my data the underlying "PhaseCorrelation" algorithm finds the drift better in edge filtered images than in the raw data)
• added option: if a ROI is put on the image, only this part of the image is considered for drift computation; the ROI is moved along with the detected drift thereby tracking the structure of interest and keeping it stable in the corrected time-series.

[ctrueden]

Thanks @tischi, this is awesome. The only issue is the change of org.scijava.vecmath back to javax.vecmath. I will merge, and then push a change back to the org.scijava vecmath again. And then I will upload this improved script to the Java-8 update site.

[ctrueden]

I uploaded the new version.

I did some brief tests, and noticed that when no ROI is present on a 2D time series, the script now fails with:

ROI is at
Traceback (most recent call last):
  File "/Applications/Science/Fiji-Java-8.app/plugins/Scripts/Plugins/Registration/Correct_3D_drift.py", line 599, in <module>
    run()
  File "/Applications/Science/Fiji-Java-8.app/plugins/Scripts/Plugins/Registration/Correct_3D_drift.py", line 553, in run
    dt = 1; shifts = compute_and_update_frame_translations_dt(imp, channel, dt, process)
  File "/Applications/Science/Fiji-Java-8.app/plugins/Scripts/Plugins/Registration/Correct_3D_drift.py", line 141, in compute_and_update_frame_translations_dt
    print "ROI is at",roi.getBounds()   
AttributeError: 'NoneType' object has no attribute 'getBounds'

I never used this script before, but I assume this is not desired behavior? @tischi If you agree, could you please see about fixing it?

[tischi]

I found the bug and fixed it. It would have also occurred on 3D data.
I staged a new pull request.