Multiseries stitching using coordinates from a tile configuration file.

[ehrenfeu]

This PR adds the possibility to stitch multi-series files (.lsm, .czi, .lif, ...) using coordinates from a TileConfiguration.txt file.

I've created this upon the reply by @ctrueden on the ImageJ mailing list. As explained in my email to the list, we need this as we're having a user that acquired a lot of datasets where each channel was saved in a separate .lsm file, where only one of the channels is leading to good results when running the stitcher. So we wanted to re-use the results calculated by the stitcher for this channel on the other channels, but the current code did not allow for specifying coordinates for a multi-series file.

A couple of remarks:

• A new optional keyword multiseries is introduced for the syntax of the TileConfiguration.txt.
• As I couldn't find any reference to the second field of the tile specification syntax, I used this one to specify the series index number of the multi-series file. Let me know if you prefer another approach, I'm happy to change the code accordingly.
• It took me a while to dig through the existing code and understand it, therefore I've added some comments that might be useful for other people looking at the code later as well. Furthermore, I felt the strong need to compact the code a little bit for being able to see more of it on my screen at one time. That's why I slightly changed the coding style of the adjusted method (e.g. placed curly brackets on the same line as an if statement). If you don't like this I can revert them, but to me this was a great help!
• For the same reason I've added convenience methods like logger(), openBF() and openBFDefault() to reduce code duplication. Although I'm quite sure @ctrueden has probably a good recommendation of how to do the logging in a much nicer way...?
• The stitcher is now also writing a TileConfiguration.registered.txt in case of using coordinates from the metadata, so the configuration created there can easily be adjusted and used for using it with the new functionality.

I have reassembled my commits and tried to group them somehow logically, hoping they are easy to understand this way. I'd be more than happy to see this merged back into master, so feel free to give me your feedback!

Some manual testing was done to cover three cases: stitching using coordinates from the metadata, stitching using individual files using a tile configuration and (of course) the new feature, stitching a single multi-series file using a tile configuration file.

[ctrueden]

@StephanPreibisch What do you think?

[ehrenfeu]

Thanks for the heads up about logging, @ctrueden - sounds promising :smile:

I personally agree that sticking to IJ.log should be enough, I just didn't want to change the original behaviour too much without discussing this step with the upstream author.

[ctrueden]

@StephanPreibisch Ping!

[ctrueden]

So, I looked over this branch, and it looks great. Thanks very much for your efforts, @ehrenfeu. Actually, I have a use case for this functionality now in my lab, which is why I'd like to expedite getting it merged. :wink: But first I have a question: how do you actually go about producing a TileConfiguration.txt file for use with this feature? With single series files, you can set the fusion method to "Do not fuse images (only write TileConfiguration)". But it seems that when stitching with "Positions from file" and "Defined by image metadata", no TileConfiguration.txt is ever written regardless. (And actually, I get an eventual OutOfMemoryError when attempting to do this.)

So: how do you create your TileConfiguration.txt for use with this feature, @ehrenfeu? FWIW, my use case is that we have an OME-TIFF dataset we'd like to stitch, but with a couple of "bad" tiles we'd like to exclude from the stitching. I thought manually tweaking a tile configuration file would be a good way to achieve this...

[StephanPreibisch]

Thanks so much @ehrenfeu and @ctrueden ... I am not very good with emails the last days. It sounds great, if you tested it let's go ahead :)

[ctrueden]

Merged. We can follow up with additional commits to address anything else related that arises. @ehrenfeu I'd still like to know how you create your TileConfiguration.txt files. By hand? Or am I missing something?

[ehrenfeu]

@ctrueden just a short follow-up on your question about creating the TileConfiguration.txt files:

• Currently I'm creating them by hand, using the logging output of the "initial" stitcher run as a basis.
• Now as this branch has been merged, I'd like to put it on my agenda to update the current methods for writing those files using the extended multiseries-syntax. I'm already having multiple users requesting this (surprise, surprise...).

[ctrueden]

Thanks for the info, @ehrenfeu.

I'd like to put it on my agenda to update the current methods for writing those files using the extended multiseries-syntax. I'm already having multiple users requesting this (surprise, surprise...).

That would be fantastic!

[iwbh15]

@ctrueden : Creating TileConfiguration.txt files I have created those files from microManager aquisitions and other metadata. Pls contact me if I can help.