Skimage.morphology.h_maxima does the result right?

why h_maxima lost some peak?

compair.rar (320.9 KB)

This is a known bug, see here:

A fix would be most welcome! =)