Test other image formats

[jakirkham]

From @jakirkham on June 7, 2017 20:19

Currently we have some tests for TIFFs. However there are no tests for any other image formats. Might be worth covering a couple basic ones. Though don't want to go too far into the weeds. Might be best to see some use cases before digging in.

Edit: Also might need to switch to something like imageio for testing so as to generate a large variety of test data.

Copied from original issue: dask-image/dask-imread#8

[jakirkham]

From @mrocklin on June 8, 2017 16:32

I don't personally have any experience here. I think that the dask.array.image.imread function defers this decision to skimage.imread. It might be worth engaging skimage devs like @jni and @stefanv with the various projects that are part of this github organization. I suspect that they are much better informed than I am.

[jakirkham]

For context, we are using PIMS under the hood, which supports a wide range of formats. Based on discussions with colleagues here, they cover many interesting cases. Out of that list we are only testing against TIFF, which is ubiquitous here. Some of the microscope formats are also interesting, but not currently tested.

Though we are only one segment of the larger image processing community. So if there are important things not in that list or on that list that need more coverage, we should discuss. Also if this turns out we need a large list of different image formats (as could well be the case), we might want to discuss a plugin architecture likely leveraging PIMS' plugin architecture.

[jakirkham]

From @stefanv on June 13, 2017 14:28

I'm not sure I understand the issue described here.

But, fwiw, I would recommend imageio, which is the backend scikit-image is switching to (after working with them to sort out licensing issues around freeimage distribution).

[jakirkham]

Unfortunately imageio alone doesn't work. The reason being we need a way to determine the shape and dtype of the underlying data without reading it all into memory. ( imageio/imageio#263 ) Once imageio addresses that, it might be possible to use it correctly.

That said, we use PIMS, which already provides us the shape and dtype info without reading full files into memory. Also PIMS provides support for using imageio amongst other image and video readers. Plus PIMS provides support for a variety of other microscope image formats that are of interest to people here that are not, to my knowledge, covered by imageio.