Question about image size and metadata

[NichD]

Greetings - I've got a project where I'm using pylibCZIrw to extract metadata and read image arrays from a CZI file containing a single image (i.e. single scene) of a stained tissue section. This is the entry point to an in-house data processing and file conversion tool chain written in python.

When looking at the extracted XML (xml_str = czidoc.raw_metadata) and the extracted metadata dict (md_dict = czidoc.metdata), I see consistent size/shape string-parameters for the stored image array:

md_dict['ImageDocument']['Metadata']['Information']['Image']['SizeX']
'15075'
md_dict['ImageDocument']['Metadata']['Information']['Image']['SizeY']
'20192'

However, when I read in the data from the first (and only) scene, I get back a data array that does not have the same size/shape as indicated in the metadata:

data_arr = czidoc.read()
print(data_arr.shape)
(20243, 15076, 3)

The shape of data_arr is consistent with info returned using the total_bounding_X methods, but I'm confused as to why these two sources of metadata differ. Additionally, I use the extracted metadata dict in a previous step for preprocessing and data-container allocation.

@FelixS90 the pypi page (https://pypi.org/project/pylibCZIrw/) indicates you're a maintainer of the python bindings for libCZI. Perhaps you can shed some light on this? Is there another place in the metadata I should be looking to determine the data array sizes?

Thanks!

NichD

[ptahmose]

I can't comment on the pylibCZIrw-behavior in this respect, but on libCZI-level the situation is something like:

• First of all, what is found in the extracted XML is "metadata", i. e. description of the data (found in the CZI), so the fields "SizeX"/"SizeY" are what they are (or what the creator of the document has put into them).
• The information retrieved by ISubBlockRepository::GetStatistics is derived from "data" (not "metadata"), and if you check the structure there are results for "two completely reasonable ways of calculating the bounding-box (which nevertheless give slightly different results)". Please see also here.

So, something along thoses lines would be what I'd suspect accounts for the discrepancy.

[NichD]

Thanks @ptahmose. Based on your response, I think I'll be more cautious about the content in metadata.