What’s New in Photutils 2.3?#

Here we highlight some of the new functionality of the 2.3 release. In addition to these changes, Photutils 2.3 includes several smaller improvements and bug fixes, which are described in the full Changelog.

Dependency Version Updates #

Photutils 2.3 bumps the minimum required versions of several key dependencies to provide users with access to the latest features and performance improvements:

NumPy minimum version is now 1.25
SciPy minimum version is now 1.11.1
Matplotlib minimum version is now 3.8
scikit-image minimum version is now 0.21

PSF Photometry Improvements #

Enhanced Error Handling and Validation #

The PSF photometry classes have been significantly improved with better error handling and validation:

These classes no longer fail for invalid sources (those with no overlap with the input data, completely masked sources, or sources with too few unmasked pixels for fitting). Instead, they assign specific flag values (64, 128, 256) to identify these conditions.
PSFPhotometry and IterativePSFPhotometry now raise an error if the input error array contains non-finite or zero values, preventing silent failures during fitting.
A new group_warning_threshold keyword has been added to control when warnings are issued about source grouping.

Improved Performance and Memory Usage #

When using Astropy 7.0+, the fitter object in PSFPhotometry and IterativePSFPhotometry now modifies the PSF model in place instead of creating copies, resulting in improved performance and significantly reduced memory usage.
For source groups, the PSF photometry classes now use a dynamically generated “flat model” approach instead of deeply nested compound models. This eliminates recursion limits that could cause failures with groups containing hundreds of sources, while providing dramatic performance improvements (up to 10x faster model creation and 50x less memory usage for large groups).

New Reduced Chi-squared Fit Statistic #

The PSF photometry classes now return a reduced chi-squared statistic (reduced_chi2 column) in the results table.

New Methods #

The PSF photometry classes now include new methods for flexible output formatting:

results_to_init_params: Converts fit results to initialization parameter format
results_to_model_params: Converts fit results to model parameter format

Enhanced Flexibility #

GriddedPSFModel can now be used with a single input ePSF model, making it equivalent to ImagePSF for simpler use cases.
The finder callable input to PSFPhotometry and IterativePSFPhotometry now can return more flexible column names beyond the previously required 'xcentroid' and 'ycentroid'.

New Utility Functions #

decode_psf_flags: A utility function for decoding PSF photometry bit flags.
PSF_FLAGS object: Provides readable, named constants for each PSF photometry bit flag and includes helper utilities for decoding.

Isophote Package Improvements #

The build_ellipse_model function in the isophote module has been Cythonized, making it considerably faster.
The build_ellipse_model function now includes an optional sma_interval keyword argument that was previously hardcoded.

Segmentation Enhancements #

`SegmentationImage` Improvements #

The to_regions method now supports a group keyword for better control over region grouping.
The polygons attribute can now handle non-contiguous segments, returning either Shapely Polygon or MultiPolygon objects as appropriate.
The to_patches and plot_patches methods now return matplotlib.patches.PathPatch objects for improved rendering.
The to_regions method now stores segment labels in the region object’s meta dictionary for easy identification.

`SourceCatalog` Features #

The make_cutouts method now includes an optional array keyword for more flexible cutout generation.

The PSFPhotometry fit_param attribute is now deprecated. Use the new results_to_init_params method instead.
The deprecated PSFPhotometry fit_results attribute has been removed.