Changelog

0.7.0.dev0 (continuous)

• No changes yet

0.6.1 / 2024-04-17

This is a no-change release to hopefully fix our zenodo integration.

0.6.0 / 2024-04-17

This version now supports Python up to 3.12 and requires at least Python 3.9. We are now using hatchling to build the package.

Features

• Fourier upsampling is now implemented for calculating the ‘refineds’ peak positions in correlation UDFs. This approach can give more precise results when peak shifts are sub-pixel at the expense of increased computation, and is available using the upsample=True argument to UDF and associated functions (see #39, #70).

• All correlation UDFs now support GPU processing with cupy, in addition to sparse input with the Fast and Sparse correlation UDFs. Conversion from unsupported backends is automatically handled using the sparseconverter package. (see #61)

Misc

• The libertem_blobfinder.common module and associated tests have been refactored to remove any dependency on LiberTEM. notably this includes all mask-generating functions previously found in libertem.masks, which are now found also in libertem_blobfinder.base.masks. As part of this change the common extra dependency group has been removed (#87).

• Move gridmatching and fullmatch from libertem.analysis to libertem_blobfinder.common.gridmatching and libertem_blobfinder.common.fullmatch since they make more sense here (#83).

0.5.0 / 2023-05-08

This version now supports Python up to version 3.11, and drops support for Python 3.6.

Features

• Integration of peak intensities at per-frame positions using libertem_blobfinder.udf.integration.IntegrationUDF. This can be used if peaks are shifted so much that integration at equal positions for all frames using libertem_blobfinder.common.patterns.feature_vector() doesn’t work anymore. (#27)

• The run_* functions that wrap various UDFs now support the progress argument of libertem.api.Context.run_udf() that was introduced in LiberTEM 0.5.0.dev0. (#22)

• Allow specifying per-frame origin shift (#23)

  • Introduce the zero_shift parameter to libertem_blobfinder.udf.correlation.CorrelationUDF, libertem_blobfinder.udf.correlation.run_fastcorrelation() and libertem_blobfinder.udf.refinement.run_refine() to specify a per-frame shift of the zero order position. This allows processing data with strong descan error.

0.4.1

Bug fixes

• Ensure compatibility with numba>=0.50 #28

0.4.0

The first release independent of LiberTEM. Changes relative to LiberTEM version 0.3.0:

Bug fixes

• Fix bounds checking, size and index calculation bugs (https://github.com/LiberTEM/LiberTEM/issues/539, https://github.com/LiberTEM/LiberTEM/pull/548)

Features

• Access to the correlation pattern for peak finding through get_correlation() (https://github.com/LiberTEM/LiberTEM/pull/571)

Misc

• The code was extracted from LiberTEM and restructured. See #1, #14, #16 and Reference for details on the new structure!

Obsolescence

• Importing blobfinder from LiberTEM is deprecated and will only be supported until 0.6.0. Import from this new package instead.