function varargout = hikmeans(varargin)
% VL_HIKMEANS  Hierachical integer K-means
%   [TREE,ASGN] = VL_HIKMEANS(DATA,K,NLEAVES) applies recursive integer
%   K-menas to cluster the data DATA, returing a structure TREE
%   representing the clusters and a vector ASGN with the data to
%   cluster assignments. The depth of the recursive partition is
%   computed so that at least NLEAVES are generated.
%
%   VL_HIKMEANS() is built on top of VL_IKMEANS() and requires the data to
%   be of class UINT8.
%
%   TREE is a structure representing the hierarchical clusters.  Each
%   node of the tree is also a structure with fields:
%
%   DEPTH::
%     Depth of the tree (only at the root node)
%
%   CENTERS::
%     K cluster centers
%
%   SUB::
%     Array of K node structures representing subtrees
%     (this field is missing at leaves).
%
%   ASGN is a matrix with one column per datum and height equal to the
%   depth of the tree. Each column encodes the branch of the tree that
%   correspond to each datum.
%
%   Example::
%     ASGN(:,7) = [1 5 3] means that the tree as depth equal to 3 and
%     that the datum X(:,7) corresponds to the branch
%     ROOT->SUB(1)->SUB(5)->SUB(3).
%
%   See also: VL_HIKMEANSPUSH(), VL_HIKMEANSHIST(), VL_IKMEANS(), VL_HELP().
[varargout{1:nargout}] = vl_hikmeans(varargin{:});