Rotate images / thumbnails in OMERO

Hi,

On a batch of slides the image of the label was taken upside-down. The label images were rotated in the acquisition software but this appears to be a transformation that is not recognised when bioformats reads the file in. As a result the labels are all upside down.

How would I go about rotating these images / thumbnails using the OMERO API?

  • It looks like the PixelsService can change the rendering settings but I don’t think a rotation is classed as a rendering setting.
  • Perhaps through the RawPixelStore and doing some getRow / putRow operations?

Thanks,

Chris

Hi Chris,

A potential fix would be to get Bio-Formats to understand the rotation at import.

We have thought about adding “rotation” and other transforms as metadata on an image (see https://trello.com/c/TkI1ih40/401-model-image-transformations) but this is a longer-term issue and I can’t give any timetable for that.

I’m afraid you can’t update the pixel data on the images themselves in OMERO.
You could generated new rotated images to replace those, e.g. using something similar to the utils/Channel_Offsets.py script. Or even Batch_Image_Export to tiff, rotate and re-import.
To “replace” the label images in the Dataset you’d have to delete them, but I expect that they are part of the same Fileset as the bigger images? In which case you can’t delete them, but would have to remove them from the Dataset.

If it’s just the Thumbnails you care about, this sounds like a case of ‘custom thumbnail’ feature request, https://trello.com/c/V5Clm9C1/24-attach-image-as-thumbnail.

Apologies there’s not a better solution right now.
Let us know if you need any more help with this,

Will

Hi Will,

thanks for the explanation.

Yes, there part of a file set of scanned slide with has a label, a macro and number of scenes. The fillesets are 1-5GB in size and there ~100 of them so I’m not keen to try this out.

I image this would be low on the priorities for BioFormats so I think for the moment we just need to make sure the labels are scanned the right way when acquiring.

Cheers,

Chris

Hi Chris,

I understand you don’t want to delete anything (you can’t actually delete part of a fileset anyway) but removing the ‘label’ images from the Dataset is a very light-weight option. You just need to decide if you’re going to put them in another Dataset or leave as “Orphaned” images.

But yes, if you can scan the labels the right way that’s the easiest option.

Cheers,

Will.