Handling Image Transformations (Scale, Rotate, Translate etc.)

Hello,

Is there a way in the current OME data model to store transformations of the image? This is important for overlaying multi-modal data (among other things, I imagine) and I am wondering if there is a standard way to store a model matrix or something similar in the OMEXML (or similar standardized metadata). If not, is that something that can be part of the #ome-ngff?

For reference, we use a custom solution here to scale/translate, but it would be nice to have something standardized (like a matrix that can be applied):

Ilan

cc: @manzt

3 Likes

@ilan-gold

The ome model supports transformation. We use a matrix to do so see


Below is an example of an ome-xml file with ROI and transformation https://downloads.openmicroscopy.org/images/OME-XML/2016-06/ROI.ome.xml

We plan to follow the same approach for #ome-ngff

Jmarie

Hi @j.burel,

Thanks so much for the pointer.

Is there any support (planned or otherwise) for 3D transformations, so an A23 entry for example?

Ilan

Hi @ilan-gold
I should clarify that the transformation in the current ome model applies only to the ROI object not the Image object.
The current ome-model only supports 2D, when switching to #ome-ngff we intend to add support for 3D (we have not started on that) i.e. the ome-model will be a “subset” of the new model.

Jmarie

1 Like

Hi @j.burel,

Thanks for the heads-up. I’m trying to figure out how to communicate this information best for an actual image, and I am not sure how. Perhaps just a separate auxiliary file for the time being will have to do. I don’t want to write a “structured annotations” parser so I may be stuck with that.

Ilan

Hi @ilan-gold ,
You may also be interested in the parallel and linked activity led by @Caterina to define an extended metadata structure (effectively an extension of OME-XML) to accompany #ome-ngff . See: Metadata for OME Next Generation File Format (NGFF) . In there we are attempting to codify exactly what you’re asking for: a set of 3D transforms (in the form of 4x4 affine matrices) to map each image to a defined world coordinate system. Would be great if you can join that and make sure your input is used and your needs are met.
Cheers,
Damir

1 Like