TODO.txt

Remember to list any API changes below in `doc/source/api_changes.txt`.

Version 0.16
------------

* Update RELEASE.txt regarding Azure Pipelines after making changes to
  scikit-image-wheels repo.
* Finalize ``skimage.future.graph`` API.
* Finalize ``skimage.future.manual_segmentation`` API.
* Remove `flatten` for `imread` in ``skimage.io._io.py``.
* In ``skimage.measure._regionprops``, remove all references to
  ``coordinates=``, ``_xycoordinates``, and ``_use_xy_warning``.
* In ``skimage.measure.moments_central``, remove ``cc`` and ``**kwargs``
  arguments.
* In ``skimage.filters.median``, change default from ``behavior='rank'`` to
  ``behavior='ndimage'``.

Version 0.17
------------

* In ``skimage.filters.median``, remove the parameters ``mask``, ``shift_x``,
  and ``shift_y``.
* Remove deprecated arguments ``beta1`` and ``beta2`` from function ``skimage.filters.frangi``
* Remove deprecated arguments ``beta1`` and ``beta2`` from function ``skimage.filters.hessian``

Other
-----
* Remove legacy pretty printing workaround for ``pytest`` in ``conftest.py``
  once minimal required ``numpy`` is set to >= 1.14.0.
* Remove deprecated ``Hxx, Hxy, Hyy`` API of ``hessian_matrix_eigvals`` in
  ``skimage.feature.corner``.
* Remove the conditional warning logic when ``numpy`` is set to >= 1.15.0
  for ``scipy`` and ``pywl`` (``pywavelet``) in ``test_lpi_filter.py`` and
  ``test_denoise.py`` regarding multidimensional indexing.
* Remove the conditional warning logic when ``numpy`` is set to >= 1.15.0
  for sparse matricies (``np.matrix``) used by scipy.sparse in
  ``test_random_walker.py``. Note that there is also a chance ``scipy.sparse``
  moves away from using ``np.matrix`` in a future version too.
* Remove the conditional import and function call when ``numpy`` is set
  to > 1.15.x in ``skimage/util/arraycrop.py``.
* When minimum supported version of ``scipy`` reaches 0.18, use 
  ``scipy.fftpack.next_fast_len`` directly in 
  ``skimage.feature.masked_register_translation``.
* Monitor SciPy for a new release that passes PYTHONOPTIMIZE builds
  `https://github.com/scipy/scipy/pull/9373`. This was identified as an issue
  in SciPy 1.1.0. When SciPy becomes compatible with PYTHONOPTIMIZE, add an
  environment variable ``PYTHONOPTIMIZE=2`` to one of the builds.
* Remove custom fused type in ``skimage/filters/_max_tree.pyx`` once
  #3486 fused types is merged.
* Check whether imread wheels are available, then re-enable testing imread
  on macOS. See https://github.com/scikit-image/scikit-image/pull/3898