Hierarchical Clustering from ICA Mixtures - On Statistical Pattern Recognition in Independent Component Analysis Mixture Modelling - page 100

Fig. 4.12 Example of image segmentation—natural and text subimages

this type of organization, the more separated hierarchical levels are, the more
different the objects are. The experiments with natural images demonstrated
meaningful hierarchy groupings of image zones with similar textures and borders.
Thus, image segmentation based on the content of the different zones was
obtained. The application of the procedure could be extended to unsupervised or
semi-supervised classification of images in order to discover meaningful contents
in the hierarchical levels. Thus, applications such as content-based image retrieval
in semantic spaces could be attempted.
Finally, note that the supervised-unsupervised scheme of the procedure (to
estimate the ICA parameters at the bottom of the hierarchy) would facilitate testing
and building the signal database in real-world applications.

References

1. B. Everitt, S. Landau, M. Leese, Cluster Analysis, 4th edn. (Arnold, London, 2001)
2. R. Xu, D. Wunsch, Survey of clustering algorithms. IEEE Trans. Neural Netw. 16(3), 645-
678 (2005)
3. D.T. Pham, A.A. Afify, Clustering techniques and their applications in engineering. Proc.
Inst. Mech. Eng. Part C J. Mech. Eng. Sci. 221(11), 1445-1459 (2007)
4. G. Lance, W. Williams, A general theory of classification sorting strategies. 1. Hierarchical
systems. Comput. J. 9(4), 373-380 (1967)