Feature clustering can be achieved using Watershed Clustering
(WC) [19], Hierarchical Agglomerative Clustering (HAC)
[20], Graph-based Normalized Cut (N-CUT) [21], Affinity
Propagation (K-AP) [22], K-Means (KM) [23], Mean-Shift
(MS) [24] or Dirichlet Process Mixture Model (DPMM) [25].
Watershed Clustering (WC) uses a grid over the input
features to calculate a density function based on the distance
among the features [19]. Cells with high feature similarity
are selected as clusters. WC is sensitive to the selection of
the cell size and its inaccurate selection could lead to an
overestimation of the size and number of clusters.