doc/source/release_notes/release_0.18.rst
We're happy to announce the release of scikit-image v0.18.3!
scikit-image is an image processing toolbox for SciPy that includes algorithms for segmentation, geometric transformations, color space manipulation, analysis, filtering, morphology, feature detection, and more.
This is a small bugfix release for compatibility with Pooch 1.5 and SciPy 1.7.
scipy.linalg.pinv2 in random_walker when
using the multigrid solver. (#5531, backport of #5437)David Manthey Gregory Lee Mark Harfouche
Gregory Lee Juan Nunez-Iglesias Mark Harfouche
We're happy to announce the release of scikit-image v0.18.2!
scikit-image is an image processing toolbox for SciPy that includes algorithms for segmentation, geometric transformations, color space manipulation, analysis, filtering, morphology, feature detection, and more.
This release mostly serves to add wheels for the aarch64 architecture; it also fixes a couple of minor bugs.
For more information, examples, and documentation, please visit our website:
master to main throughout (#5243, #5295)This is a bug fix release and contains the following two bug fixes:
See below for the new features and API changes in 0.18.0.
We're happy to announce the release of scikit-image v0.18.0!
scikit-image is an image processing toolbox for SciPy that includes algorithms for segmentation, geometric transformations, color space manipulation, analysis, filtering, morphology, feature detection, and more.
This release of scikit-image drops support for Python 3.6 in accordance with
the NEP-29 Python and Numpy version support community standard <https://numpy.org/neps/nep-0029-deprecation_policy.html>_: Python 3.7 or
newer is required to run this version.
For more information, examples, and documentation, please visit our website:
skimage.filters.rank.skimage.future. Try it out! (#4739)segmentation.expand_labels to dilate labels while
preventing overlap (#4795)measure.regionprops and regionprops_table (#4810)data subpackage: data.eagle
(#4922), data.human_mitosis (#4939), data.cells3d (#4951), and
data.vortex (#5041). Also note that the image for data.camera has
been changed due to copyright issues (#4913).skimage.feature.structure_tensor now supports 3D (and nD) images as input
(#5002)measure.regionprops now supports multichannel intensity images (#5037)transform.AffineTransform (#4733)visualizing 3D data <https://scikit-image.org/docs/dev/auto_examples/applications/plot_3d_image_processing.html>_ (#4850)rgb2lab (#4839) and marching_cubes (#4846)measure.marching_cubes, mentioning how to decimate a
mesh using mayavi (#4846)util.random_noise (#5001)morphology.h_maxima and morphology.h_minima
(#4929).util.img_as_int (#4888).how to download example datasets <https://scikit-image.org/docs/dev/user_guide/install.html#downloading-all-demo-datasets>_
which are not installed with scikit-image (#4984). Similarly, the contributor
guide has been updated to mention how to host new datasets in a gitlab
repository (#4892).benchmarking section of the developer documentation <https://scikit-image.org/docs/dev/development/contribute.html#benchmarks>_
has been expanded (#4905).pyramid_gaussian (#4696), Richardson-Lucy deconvolution (#4880)skimage.restoration.richardson_lucy, computations are now done in
single-precision when the input image is single-precision. This can give a
substantial performance improvement when working with single precision data.filter_epsilon keyword argument
to avoid division by very small numbers (#4823)measure.find_contours (#4862)skimage.segmentation.slice) was improved for the case where a mask
is supplied by the user (#4903). The specific superpixels produced by
masked SLIC will not be identical to those produced by prior releases.exposure.adjust_gamma has been accelerated for uint8 images by using
a look-up table (LUT) (#4966).measure.label has been accelerated for boolean input images, by using
scipy.ndimage's implementation for this case (#4945).util.apply_parallel now works with multichannel data (#4927).skimage.feature.peak_local_max supports now any Minkowski distance.morphology.convex_hull_image now uses much less memory by checking hull
inequalities in sequence (#5020)skimage.restoration.richardson_lucy returns a single-precision output
when the input is single-precision. Prior to this release, double-precision
was always used. (#4880)threshold_rel in skimage.feature.corner has
changed from 0.1 to None, which corresponds to letting
skimage.feature.peak_local_max decide on the default. This is currently
equivalent to threshold_rel=0.measure.label, the deprecated neighbors parameter has been
removed. (#4942)data.camera has changed because of copyright
issues (#4913).label2rgb has been fixed when the input image had np.uint8
dtype (#4661)skimage.color.separate_stains (#4725)peak_local_max (#2592, #4756, #4760,
#5047)random_walker when input labels have negative values (#4771)bg_color parameter in
label2rgb when its value is a string (#4840)metrics.variation_of_information
(#4875)skimage.measure.regionprops was erroneous
for 3D objects, since it did not take tunnels into account. A new implementation
based on integral geometry fixes this bug (#4380).skimage.morphology.selem.rectangle the height argument
controlled the width and the width argument controlled the height.
They have been replaced with nrow and ncol. (#4906)skimage.segmentation.flood_fill and skimage.segmentation.flood
now consistently handle negative values for seed_point.segmentation.flood have been fixed (#4948, #4972)draw.polygon for the case of 0-d input has been fixed
(#4943).registration.phase_cross_correlation, a ValueError is raised when
NaNs are found in the computation (as a result of NaNs in input images).
Before this fix, an incorrect value could be returned where the input images
had NaNs (#4886).indices argument in skimage.feature.peak_local_max has been
deprecated. Indices will always be returned. (#4752)skimage.feature.structure_tensor, an order argument has been
introduced which will default to 'rc' starting in version 0.20. (#4841)skimage.feature.structure_tensor_eigvals has been deprecated and will be
removed in version 0.20. Use skimage.feature.structure_tensor_eigenvalues
instead.skimage.viewer subpackage and the skivi script have been
deprecated and will be removed in version 0.20. For interactive visualization
we recommend using dedicated tools such as napari <https://napari.org>_ or
plotly <https://plotly.com>_. In a similar vein, the qt and skivi
plugins of skimage.io have been deprecated
and will be removed in version 0.20. (#4941, #4954)skimage.morphology.selem.rectangle the arguments width and
height have been deprecated. Use nrow and ncol instead.threshold_rel=0` was removed from the Examples of the following docstrings: skimage.feature.BRIEF, skimage.feature.corner_harris, skimage.feature.corner_shi_tomasi, skimage.feature.corner_foerstner, skimage.feature.corner_fast, skimage.feature.corner_subpix, skimage.feature.corner_peaks, skimage.feature.corner_orientations, and skimage.feature._detect_octave``.skimage.restoration._denoise, the warning regarding
rescale_sigma=None was removed.skimage.restoration._cycle_spin, the # doctest: +SKIP was removed.pyproject.toml has been added to the sdist.myst (#4863).io.*.py (#4845)test_ functions (#5044)indices argument from examples (#5082)A warm thank you to all contributors who added to this release. A fraction of contributors were first-time contributors to open source and a much larger fraction first-time contributors to scikit-image. It's a great feeling for maintainers to welcome new contributors, and the diversity of scikit-image contributors is surely a big strength of the package.