search

scifio / scifio-ome-xml

SCIFIO plug-in providing support for the OME data model, including readers, writers and translators for OME-XML and OME-TIFF formats.
http://scif.io/
BSD 2-Clause "Simplified" License
2 stars 3 forks source link

Ensure Plane DeltaT values are stored into SCIFIO axes #17

Open hinerm opened 10 years ago

hinerm commented 10 years ago

Depends on https://github.com/imglib/imglib/issues/64

If there are Plane DeltaT values present in OME-XML, we should ensure the Axes of the parsed/translated metadata store them appropriately.

stelfrich commented 8 years ago

@ctrueden's comment ("non-linear SCIFIO axes to store the time stamps in a data-model-agnostic way") in https://github.com/fiji/fiji/issues/74 got me interested in that topic.

I dug a little into the code and stumbled across the ImageMetadata API that supports only one axis per Image. Correct me if I am wrong but doesn't that assume that two Planes from different channels have been acquired at the same time since they can be only mapped to one time-Axis?

If that topic is off the table, feel free to ignore my comment :smile:

ctrueden commented 8 years ago

Whoa, how has this PR been sitting for over two years? Because the upstream imglib2 PR has sat that long? Maybe we can resolve this in the upcoming data model hackathon, as part of our work on the RichImage.

@stelfrich You're right: the fact that every Plane in OME-XML can have a different DeltaT is more flexible than ImageJ2's axis-driven metadata model. The original "variable width time stamper" would not require the more flexible sort of metadata—just a non-linear time axis. But if we want to fully support what different file formats are already storing sometimes, I guess we need to make the metadata model more flexible. I fear the complexity but it seems like the correct thing to do.

stelfrich commented 8 years ago

Maybe we can resolve this in the upcoming data model hackathon, as part of our work on the RichImage.

That was @dietzc's idea as well.

I fear the complexity [...]

We just have to find a good balance between complexity and usability (as always). The use case that got me interested is actually pretty simple: label each slice of an image sequence with time stamps where the first time point shall be the DeltaT of that Image (speaking in OME-XML terms). While it might be "easy" to extend the metadata model to support that case (just using one enumerating axis per Image), the image sequence is put in the broader context of an experiment which is not as easy to determine/model..

Another use case that crossed my mind is a rather common setup in live cell fluorescence imaging: acquire a widefield image every 5 minutes and a fluorescence image every 20 minutes. The acquired image sequence is one Image with the two channels having different Axis. That would be the most obvious example to start thinking about a more complex metadata model..