Driftcorrection does not allow for rotation across recording epochs

[stevevanhooser]

We need to enhance our drift correction procedures to allow for a rotation across recording epochs. This is necessary because we are now using small screw-on microscopes that change orientation from recording session to recording session.

[stevevanhooser]

How to do it?

It is a little tricky because right now our drift correction only assumes linear shifts in X and Y.

Further, these image sets will be taken across different sessions, so they may include different cells, different views, etc.

Option 1) Run a separate little program to find the best rotation and X / Y shift and re-generate the TIFF files. Would need a "write frame" function.

Option 2) Update all of our drift correction code to optionally include rotation

I like option 1 better. If we find an approach in option 1 that is successful, we can think about implementing option 2. If we stink at the process, then we wouldn't have updated a whole bunch of code for no benefit

[stevevanhooser]

OK, how to do it?

Always, there will be a set of sets of recording epochs.

Let's call the set of epochs at a given orientation EO.

Then all epochs are

E = { EO1, EO2, ..., EOn}

This list is known to the user and needs to be specified to the program.

There will be a few situations. In some situations, EOi and EOj will be nearly identical up to a rotation and shift, and the user will want to attempt to indicate that these epochs are recordings of the same tissue. In other situations, EOi and EOj will be similar but not exactly identical. The user may still want to rotate EOj to match EOi. Finally, there may be other situations where the quality of the match is unclear; the user may want to try to align the epochs exactly but may need to see if it is possible.