dask-image 0.4.0

0.4.0 (2020-09-02)

We’re pleased to announce the release of dask-image 0.4.0!

Highlights

The major highlight of this release is support for cupy GPU arrays for dask-image subpackages imread and ndfilters. Cupy version 7.7.0 or higher is required to use this functionality. GPU support for the remaining dask-image subpackages (ndmorph, ndfourier, and ndmeasure) will be rolled out at a later date, beginning with ndmorph.

We also have a new function, threshold_local, similar to the scikit-image local threshold function.

Lastly, we’ve made more improvements to the user documentation, which includes work by new contributor @abhisht51.

New Features

• GPU support for ndfilters & imread modules (#151)

• threshold_local function for dask-image ndfilters (#112)

Improvements

• Add function coverage table to the dask-image docs (#155)

• Developer documentation: release guide (#142)

• Use tifffile for testing instead of scikit-image (#145)

3 authors added to this release (alphabetical)

• Abhisht Singh - @abhisht51

• Genevieve Buckley - @GenevieveBuckley

• jakirkham - @jakirkham

2 reviewers added to this release (alphabetical)

• Genevieve Buckley - @GenevieveBuckley

• Juan Nunez-Iglesias - @jni

0.3.0 (2020-06-06)

We’re pleased to announce the release of dask-image 0.3.0!

Highlights

• Python 3.8 is now supported (#131)

• Support for Python 2.7 and 3.5 has been dropped (#119) (#131)

• We have a dask-image quickstart guide (#108), available from the dask examples page: https://examples.dask.org/applications/image-processing.html

New Features

• Distributed labeling has been implemented (#94)

• Area measurement function added to dask_image.ndmeasure (#115)

Improvements

• Optimize out first where in label (#102)

Bug Fixes

• Bugfix in center_of_mass to correctly handle integer input arrays (#122)

• Test float cast in _norm_args (#105)

• Handle Dask’s renaming of atop to blockwise (#98)

API Changes

• Rename the input argument to image in the ndimage functions (#117)

• Rename labels in ndmeasure function arguments (#126)

Support

• Update installation instructions so conda is the preferred method (#88)

• Add Python 3.7 to Travis CI (#89)

• Add instructions for building docs with sphinx to CONTRIBUTING.rst (#90)

• Sort Python 3.7 requirements (#91)

• Use double equals for exact package versions (#92)

• Use flake8 (#93)

• Note Python 3.7 support (#95)

• Fix the Travis MacOS builds (update XCode to version 9.4 and use matplotlib ‘Agg’ backend) (#113)

7 authors added to this release (alphabetical)

• Amir Khalighi - @akhalighi

• Elliana May - @Mause

• Genevieve Buckley - @GenevieveBuckley

• jakirkham - @jakirkham

• Jaromir Latal - @jermenkoo

• Juan Nunez-Iglesias - @jni

• timbo8 - @timbo8

2 reviewers added to this release (alphabetical)

• Genevieve Buckley - @GenevieveBuckley

• jakirkham - @jakirkham

0.2.0 (2018-10-10)

• Construct separate label masks in labeled_comprehension (#82)

• Use full to construct 1-D NumPy array (#83)

• Use NumPy’s ndindex in labeled_comprehension (#81)

• Cleanup test_labeled_comprehension_struct (#80)

• Use 1-D structured array fields for position-based kernels in ndmeasure (#79)

• Rewrite center_of_mass using labeled_comprehension (#78)

• Adjust extrema’s internal structured type handling (#77)

• Test labeled_comprehension with object type (#76)

• Rewrite histogram to use labeled_comprehension (#75)

• Use labeled_comprehension directly in more function in ndmeasure (#74)

• Update mean’s variables to match other functions (#73)

• Consolidate summation in _ravel_shape_indices (#72)

• Update HISTORY for 0.1.2 release (#71)

• Bump dask-sphinx-theme to 1.1.0 (#70)

0.1.2 (2018-09-17)

• Ensure labeled_comprehension’s default is 1D. (#69)

• Bump dask-sphinx-theme to 1.0.5. (#68)

• Use nout=2 in ndmeasure’s label. (#67)

• Use custom kernel for extrema. (#61)

• Handle structured dtype in labeled_comprehension. (#66)

• Fixes for _unravel_index. (#65)

• Bump dask-sphinx-theme to 1.0.4. (#64)

• Unwrap some lines. (#63)

• Use dask-sphinx-theme. (#62)

• Refactor out _unravel_index function. (#60)

• Divide sigma by -2. (#59)

• Use Python 3’s definition of division in Python 2. (#58)

• Force dtype of prod in _ravel_shape_indices. (#57)

• Drop vendored compatibility code. (#54)

• Drop vendored copy of indices and uses thereof. (#56)

• Drop duplicate utility tests from ndmorph. (#55)

• Refactor utility module for imread. (#53)

• Reuse ndfilter utility function in ndmorph. (#52)

• Cleanup freq_grid_i construction in _get_freq_grid. (#51)

• Use shared Python 2/3 compatibility module. (#50)

• Consolidate Python 2/3 compatibility code. (#49)

• Refactor Python 2/3 compatibility from imread. (#48)

• Perform 2 * pi first in _get_ang_freq_grid. (#47)

• Ensure J is negated first in fourier_shift. (#46)

• Breakout common changes in fourier_gaussian. (#45)

• Use conda-forge badge. (#44)

0.1.1 (2018-08-31)

• Fix a bug in an ndmeasure test of an internal function.

0.1.0 (2018-08-31)

• First release on PyPI.

• Pulls in content from dask-image org.

• Supports reading of image files into Dask.

• Provides basic N-D filters with options to extend.

• Provides a few N-D Fourier filters.

• Provides a few N-D morphological filters.

• Provides a few N-D measurement functions for label images.

• Has 100% line coverage in test suite.