Choose reference slice in sequence dimension

[7rebor]

This is going to be a confusing post as the nomenclature is difficult.

I would like to choose one time slice as a reference for the sequence to align to. Can I do this already? As far as I'm aware, the reference in the sequence is always the first slice (e.g. first timepoint). It is not obvious from UI description as to whether the sequence reference (sliders) are used to limit the sequence that is transformed or to use a single slice as a reference? To me, it seems like the sliders are used to only register a small part of the sequence and transform only that part, is this correct?

Are all of the 'reference' options in the registration dimensions just for the computation? And the transformation is applied to all along the sequence/registration dimensions?

[tischi]

1. Currently the image at the min position of the sequence dimension is the reference. This cannot be change at the moment. Do you think it is an issue?
2. The transformation is applied along all dimensions! However it only is computed considering the chosen reference interval. Does that make sense? I found it complicated, too, when designing it...hard to write N-Dimensional code :-)

[7rebor]

Currently the image at the min position of the sequence dimension is the reference. This cannot be change at the moment. Do you think it is an issue?

I was going to say it is not an issue, as it will align all images of the sequence to the first image. However, if the first image is particularly offset compared to the others, all images will be translated way off origin (the image canvas size isn't increased, so essentially you might lose lots of information in all of the subsequent images).

In a scenario where you can choose your own reference (using the sliders), you could account for this by choosing the image in the sequence that is nearest to all images and hence produces the smallest overall change to all images. This could be negated by increasing the canvas size to accommodate.

However, I don't have this problem currently.

The transformation is applied along all dimensions! However it only is computed considering the chosen reference interval. Does that make sense? I found it complicated, too, when designing it...hard to write N-Dimensional code :-)

Yep, makes sense!

[tischi]

the image canvas size isn't increased, so essentially you might lose lots of information in all of the subsequent images

In fact the canvas size should be increased! If it isn't then there is a bug...
Could you check?

[7rebor]

Ah, big data viewer might increase it, didn't check. But the resultant imagej stack canvas size is not increased (for XY at least). If the image has been translated out of the canvas field of view the resultant image looks like this:

I haven't updated the plugin since a few weeks ago, was it recently updated to not alter canvas size? Or should it have always been a feature?

[tischi]

Please try updating your Fiji. It should increase the canvas size.