Masked arrays not handled by detection algorithms.

[StefanoPierini-Codra]

Problem description

Detection functions pass numpy.ma.MaskedArray directly to underlying libraries (scikit-image, OpenCV, scipy) that don't support masked arrays.

Affected functions: get_2d_peaks_coords, get_hough_circle_peaks, find_blobs_dog, find_blobs_doh, find_blobs_log, find_blobs_opencv.

Proposed solution

Partial solutions that could work in some case but not in all of them, a minimal fix would be to run the computation raising a warning that explain that the analysis is not supported ad could give unexpected results.

[StefanoPierini-Codra]

Summary

Detection functions (get_2d_peaks_coords, get_hough_circle_peaks, find_blobs_dog, find_blobs_doh, find_blobs_log, find_blobs_opencv) now emit a UserWarning when called with a masked array (e.g. an image restricted to a non-rectangular ROI). The underlying libraries (scikit-image, OpenCV, SciPy) don't support masked arrays and silently ignore the mask, so results can be unexpected inside or near masked areas — the warning makes this limitation visible instead of failing silently.

Changes

• New warn_if_masked decorator in sigima/tools/checks.py.
• Applied to all six detection functions in sigima/tools/image/detection.py.
• New dedicated test file: sigima/tests/image/detection_masked_array_unit_test.py.
• Release note added in doc/release_notes/release_1.02.md.