[AlexKurek]

Every 10-sigma island is matched against the bright-source peaks by locating its peak (np.nanargmax) and computing np.hypot against each bright peak. Although almost all islands are far from any bright source.

Fix

• Store bright-source peaks once in bright_xy
• Skip an island early if no bright peak lies within peak_match_tol pixels of its bounding box
• Compute distances with one np.hypot
• Name the 2.5 px tolerance peak_match_tol

Before:

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        5  253.784   50.757  446.185   89.237 /home/akurek/miniconda3/envs/pybdsf_3_15/lib/python3.15/site-packages/bdsf/rmsimage.py:134(__call__)

After:

        5   28.082    5.616  210.888   42.178 /home/akurek/miniconda3/envs/pybdsf_3_15/lib/python3.15/site-packages/bdsf/rmsimage.py:134(__call__)

The whole PyBDSF run should be ~10% faster.

[AlexKurek]

Im sorry, this will influence the results of wavelets. Closing.