imageio v3 api reads all pages of a multipage TIFF at once

I have a two-page TIFF, one 8bit grayscale image of shape (15,10) per page, and I am expecting imageio.v3.imiter to iterate over the two pages. Yet, it reads them both at once:

>>> from skimage import data_dir
>>> file = data_dir + '/multipage.tif'
>>> import imageio
>>> len([*imageio.v3.imiter(file)])
1
>>> for img in imageio.v3.imiter(file):
...     img.shape
... 
(2, 15, 10)

The same code prior to #805 fulfills my expectation and iterates over the two pages:

>>> from skimage import data_dir
>>> file = data_dir + '/multipage.tif'
>>> import imageio
>>> len([*imageio.v3.imiter(file)])
2
>>> for img in imageio.v3.imiter(file):
...     img.shape
... 
(15, 10)
(15, 10)

The TIFF file in question is part of skimage: https://github.com/scikit-image/scikit-image/blob/c95a8fd46a49a38de1f6032c84d9d61160b78283/skimage/data/multipage.tif It appears to have been created with ImageMagick.

[almarklein]

@FirefoxMetzger good luck with your thesis!! 😃

[lschr]

@FirefoxMetzger Thanks for the reply!

So the expected result should be that you get one ndimage instead of a sequence of frames.

I guess you are right. However, since contiguous=True used to be the default (and I wasn’t even aware of this option), I have now a lot of old files with all images in a single series, and I guess I am not the only one. This was just one example illustrating why I find it useful to be able to read individual pages. The other, more concerning one is commercial software over which we have no control writing all images into a single series.

The short answer is unfortunately no because we are moving away from the image modes iIvV?

I wasn’t thinking about permanently reverting #816, but only until there is another way to load single pages (i.e., #914 is merged). I really appreciate your work moving imageio forward!

if you want to test this early, feel free to have a look at the PR 😃 I can’t give an exact PR because I am juggling a few things in parallel right now, but I promise to incorporate feedback before it gets merged 😉

Cool, I will have a look.