MATLAB packages available below:

• Unsupervised data segmentation  
• Natural image segmentation
• Supervised classification

All code is Copyright 2007 University of Illinois. 

Clustering of multivariate mixed data

   Matlab package:   segmentation_code.zip

   The usage is simple: if  X  is the data matrix (in column form), calling
        Y = coding_seg(X, epsilon, false);
    will segment the data so as to minimize the number of bits needed to
    code the data vectors upto distortion epsilon^2. The last parameter
    should be true if the data are known to be nonzero-mean. See "test.m"
    for an example.

   Contact: John Wright (jnwright@uiuc.edu).


Classification of multivariate data 

    Matlab package:   MICL_code.zip

    Contact: John Wright (jnwright@uiuc.edu)


Segmentation of natural images

    Matlab package:   tbes_1.0.zip

    Installation: This texture segmentation algorithm uses the superpixel code of Mori et al., which includes some c files that need to be compiled. To do this open MATLAB and then run the following commands:

> cd superpixels/yu_imncut
> mex csparse.c
> mex ic.c
> mex imnb.c
> mex parmatV.c
> mex spmd_imncut.c

    Usage: To test our texture segmentation algorithm, run the following command in MATLAB:

> test_texture_seg

    Please refer to our image segmentation webpage for details and extensive results / benchmarking.

    Contact: Shankar Rao (srrao AT illinois DOT EDU).