For LIFs, img.physical_pixel_sizes.X returns px/micon and not micron/px

[psobolewskiPhD]

System and Software

• aicsimageio Version: 4.0.4
• Python Version: 3.9.6
• Operating System: MacOS BigSur 11.4 (Apple Silicon M1) (everything native arm64)

Description

Followup from:
https://forum.image.sc/t/setting-scale-bar-units-in-other-than-pixels-real-coordinates/49158/13?u=psobolewskiphd
When importing scenes from a LIF, img.physical_pixel_sizes.X which should return pixel size or a value of length (micron, nm) per pixel instead returns the reciprocal: px/length

You can access the LIF and the XML in the closed issue:
#277
Here's a relevant part of the XML:

DimID="1" NumberOfElements="1920" Origin="0.000000e+000" Length="1.244352e-003" Unit="m" BitInc="0" BytesInc="1" />
<DimensionDescription DimID="2" NumberOfElements="1440" Origin="0.000000e+000" Length="9.331016e-004" Unit="m" 

Based on this (and I've verified this with LAS X) the pixel size is:
(1.244e-3 m)/(1920 px) = 648 nm = 0.648 micron
Meanwhile:

>>> img.physical_pixel_sizes.X
1.542168132489842

which is 1/(pixel size)

Expected Behavior

That img.physical_pixel_sizes.X return the pixel size as advertised in the docs:

img.physical_pixel_sizes.X # returns the X dimension pixel size as found in the metadata

where 'size' means a value in the length dimension (micron, nm, m).

Reproduction

(You can access the LIF and the XML in the closed issue:
#277)
To set the scale in Napari, one needs to do 1/img.physical_pixel_sizes.X to get micron/pixel.

from aicsimageio import AICSImage
import napari
import dask.array as da
# if you want in-memory
import numpy as np
from vispy.color import Colormap

img = AICSImage("../Documents/Leica DMi8/20210428/20210428_24w_L929_Ho_B2C3.lif", reconstruct_mosaic=False)
scene_stack = []
for i, scene in enumerate(img.scenes):
        if i >46:
                img.set_scene(i)
                scene_stack.append(img.get_image_dask_data("YX"))
#                scene_stack.append(img.dask_data)

scene_stack = da.stack(scene_stack)
#scene_stack = np.stack(scene_stack)

#get pixel sizes (this is px/micron)
img.physical_pixel_sizes.X
microns_per_pixel = [1/img.physical_pixel_sizes.X, 1/img.physical_pixel_sizes.X]
        
I_Blueish = Colormap([[1, 1, 1], [0, .4, 1], [0, 0, 0]])
viewer = napari.view_image(scene_stack, colormap=('I Blueish', I_Blueish), scale=microns_per_pixel)
viewer.scale_bar.unit = "um"

Environment

My entire python environment is Apple Silicon M1 (arm64) native.

[toloudis]

Took a look at the documentation and code for readlif and it does indeed look like we have to use 1/scale instead of scale