0.23.1

Bug fixes

• skimage.io.imsave didn't work with cle images anymore

0.23.0

New features

• Device selection using device index by @haesleinhuepf in #258
• Warning in case a selected device name wasn't found @haesleinhuepf in #258
• enable color bars in subplots by @haesleinhuepf in #260 (thanks to @somas193 for proposing and providing code)

Bug fixes

• cle_image.T did not work as expected
• cle.create_like(numpy_array) created an image of wrong size (shape was inverse).
• iterating through slices of a 3D stack didn't work

Backwards compatibility breacking changes

• The behaviour of cle.create_like() has changed in case a numpy array was passed (see above)
• The alias local_threshold was removed. Use greater instead.

Full Changelog: 0.22.1...0.23.0

0.22.1

New features

• make excluding edges on specific edges optional by @haesleinhuepf in #248
• add generate_mean_intensity_between_points_matrix and generate_standard_deviation_intensity_between_points_matrix by @haesleinhuepf in #257

Bug fixes

• bugfix EDF result dimensionality by @haesleinhuepf in #256 (thanks to @jaimemcc for reporting)

Full Changelog: 0.22.0...0.22.1

0.22.0

Bug fixes

• Issue with indexing linux (Thanks to @ENicolay for reporting and support in fixing)
• Issue with networkx 3.0, from_numpy_matrix was deprecated and conversion failed.

Backwards compatibility breaking bug fixes

• The visualisation of label images was changed: The first two colors were flipped (now new: 1=cyan, 2=orange). The colour order is now identical with matplotlib default colours.

0.21.2

Bugfixes

• Tests failed because of numpy 1.24 release. Library code remains unchanged

0.21.1

New features

• generate_should_touch_matrix()
• merge_annotated_touching_labels()

Bug fixes

• merge_labels_according_to_touch_matrix() didn't merge all requested labels in case labels were touching in chains / circles

0.21.0

Backwards compatibility breaking changes (MacOS, Intel GPUs only)

• This version changes the default device on older MacOS because Intel based GPUs caused trouble. The CPU is now used per default. This gives users of older MacBooks a functional OpenCL device and more available memory.

0.20.0

New features

• Added maximum position function by @grahamross123 in #230
• Pytorch interoperability by @haesleinhuepf as proposed by @alisterburt in #232
• Colab example by @haesleinhuepf in #222

Bug fixes

• Bugfix cupy backend by @haesleinhuepf in #231
• Orthogonal interpolation (Sapoznik et al. (2020) https://doi.org/10.7554/eLife.57681) for deskewing by @pr4deepr with support by @dpshepherd in #224 Also thanks for reporting this bug to Sebastian Bundschuh (MPI-CBG)

Backwards compatibility breaking changes

• The parameter linear_interpolation in the functions deskew_x and deskew_y have now True as default parameter and the result image looks different because of a bugfix, now using orthogonal interpolation (see above). See this notebook for details

Deprecated functions and parameters

• The parameter linear_interpolation in the functions deskew_x and deskew_y is deprecated. It will be removed in a future version.

Full Changelog: 0.19.4...0.20.0

0.19.4

New features

• Added new columns to statistics_of_labeled_neighbors for statistics of dilated label images. (thanks to @lcferme for proposing)

Bug fixes

• The touch count and touch portion measurements in statistics_of_labeled_neighbors considered background as something that touches. This has been corrected. Background is ignored now, In case you want to measure touching with background, add 1 to the image. This makes background a valid label and allows quantifying touches with background.

0.19.3

New features

• crop_border
• draw_distance_mesh_between_n_nearest_labels
• added auto-generated docs