cgdl

->

const int T = 7; /* target clusters number */

struct mat_t *dm = mat_load("distance-matrix.txt"); 
struct cgdl_t *gdl = cgdl_new(4,1);

cgdl_init(gdl, dm);

mat_free(&dm);

while (cgdl_num(gdl) > T) {
      cgdl_merge(cgdl);
}

cgdl_free(&cgdl);

This is an implementation of the agglomerative clustering algorithm
presented in [0]. There's an example presented in example.c. It
simply grouped the pixels in the first picture using GDL, the second
picture is the output.

The codes were tested both on Win7 using VS2010 and Linux (Ubuntu 12.04), and it should work in many other systems too. You need install the OpenCV (1.0+) library to compile example.c.

On Windows: Just add all source files into a project, setup the OpenCV directories, compile.
On Linux: Type make to compile, then run the demo with ./cgdl-demo .

The author also open sourced his Matlab implementation in :
https://github.com/waynezhanghk/gactoolbox

[0]: Graph Degree Linkage: Agglomerative Clustering on a Directed Graph, Wei Zhang. ECCV 2012.