One can often see clustering in a hard form,
which assigns each pixel xi to one and only one cluster.
A soft or a fuzzy technique, on the other hand,
assigns to each pixel xi a fractional degree of membership uija[0, 1] for all clusters.
The higher the degree of the membership uij,
the more probable it is that the pixel xi belongs to cluster
j. A deterministic similarity is often used in hard
clustering and a probabilistic distance (or a fuzzy variant)
is used in soft clustering. A soft clustering contains more