Home > Academics, Research > Balanced-Tree Structure Bayesian Networks (TSBNs) for Image Segmentation

Balanced-Tree Structure Bayesian Networks (TSBNs) for Image Segmentation

In this work, I made a more generalized version of Gaussian mixture model (GMM) by putting prior balanced-tree structure over the class variables (mixture components) in the hope that the induced correlation among the hidden variables would suppress the noise in the resulting segmentation. Unlike supervised image classification by [1], this work focuses on totally unsupervised segmentation using TSBN. In this work, it is interesting to see how the data will be “self-organized” according to the initial structure given by TSBN.

The MATLAB code is available here. The codes call inference routines in Bayesian network toolbox (BNT), so you may want to install the toolbox before using my TSBN code.

original image 136x136

original image 136x136

16x16 feature image. Each pixel represents 16-dimensional vector.

16x16 feature image. Each pixel represents 16-dimensional vector.

Segmentation result using TSBN by setting predefined number of classes to 3. It turns out that the classes are meaningful as they are sky, skier and snow.

Segmentation result using TSBN by setting predefined number of classes to 3. It turns out that the classes are meaningful as they are sky, skier and snow.

[1]  X. Feng, C.K.I. Williams, S.N. Felderhof, “Combining Belief Networks and Neural Networks for Scene Segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 467-483, April, 2002

Advertisements
  1. Gif
    February 15, 2011 at 1:29 pm

    What are the image features you are using here?

    • admin
      February 15, 2011 at 1:33 pm

      The features are RGB, L*a*b, and first 10 zigzag coefficients of DCT (using dct2 in MATLAB). In fact, you can add more features of your choices, but in this preliminary experiment I just use only a handful of simple features. Good luck on you work too!

  2. Gif
    February 15, 2011 at 1:41 pm

    I found that the zigzag pattern for DCT is not easy to generate, will you suggest any helpful website showing how to generate the pattern efficiently?

  3. Gif
    February 15, 2011 at 1:46 pm

    Do you have reply e-mail notification system? So that I can know immediately when you reply. Thanks!

    • admin
      February 15, 2011 at 1:47 pm

      Good idea, Gif. I’m fixing it right now. Thanks for letting me know ^_^.

  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: