DIPlib uses a different name for the various possible connectivites than
you might be used to. This is to generalize this parameter to images of
any dimensionality. It is defined as follows: if `connectivity` is 1 all pixels
for which only one coordinate differs from the pixel's coordinates by
maximally 1 are considered neighbours; if it is 2, all pixels for which one
or two coordinates differ maximally 1 are considered neighbours.
The connectivity can never be larger than the image dimensionality.

In terms of the obsolete connectivity definitions we have:

In 2-D | this connectivity | corresponds to | and forms this structuring element |

1 | 4 connectivity | diamond | |

2 | 8 connectivity | square | |

-1 | 4-8 connectivity | octagon | |

-2 | 8-4 connectivity | octagon | |

In 3-D | this connectivity | corresponds to | and forms this structuring element |

1 | 6 connectivity | octahedron | |

2 | 18 connectivity | cuboctahedron | |

3 | 26 connectivity | cube | |

-1 | 6-26 connectivity | small rhombicuboctahedron | |

-3 | 26-6 connectivity | small rhombicuboctahedron |

The negative connectivities are only defined for the functions in binary
morphology such as `BinaryDilation` and `BinaryErosion`. These
alternate steps with different connectivity to produce a better approximation
to an isotropic structuring element.