[gsoc2019] Add tests for adaptive thresholding

[mloskot]

@miralshah365 has submitted implementation of the algorithm in #341 with example-as-test, which has worked fine for development, but we need more complete test suite for it.

We need a basic variety of input images.

References

Test images options are being discused arranged:

• PNM format is going to move to the core features to be offered as built-in, see [RFC] Managing input images for tests"
• Hosting of standard images for IO tests (and not only), see [IO] Hosting test images

[manav-yb]

Is there any work to be still done on this issue?

[mloskot]

AFAIK, yes, it an open issue and has not been targetted with any PR yet.

[AdityaSanthosh]

@mloskot So, we basically need to upload more PNM files in this repo https://github.com/chhenning/gil_io_images/tree/master/pnm by creating them using ImageMagick if we're going with Julia kind of maintaining test Images in external Repo. Is there any script to download them into my local boost gil repo? If not, what does it take to create such a script that deploys test images directly into my boost gil test folder?

[mloskot]

@AdityaSanthosh We maintain test images in this external repo under a GIL-dedicated organization https://github.com/boost-gil/test-images

Next, we need a script (i.e. two scripts, in Bash and PowerShell) that download the images from that repo into explicitly specified or automatically determined known location e.g. libs/gil/test/images).

Then, we need to

• update the CI scripts to use those download scripts
• modify tests to recognise the presence of the test images

Finally, we can get rid of any test images hosted in the Boost.GIL repo or at least any binary images and just keep the ASCII PNM but that will be possible after https://github.com/boostorg/gil/issues/359 is implemented.

---

Alternatively, we could Git submodule the test-images inside Boost.GIL repo, but I'm not convinced it is a good idea at all.