Color Image Segmentation upon a New Unsupervised Approach using Amended Competitive Hebbian Learning

Meriem Timouyas, Souad Eddarouich, Ahmed Hammouch

2016

Abstract

This paper proposes a new unsupervised color image segmentation procedure based on the competitive concept, divided into three processing stages. It begins by the estimation of the probability density function, followed by a training competitive neural network with Mahalanobis distance as an activation function. This stage allows detecting the local maxima of the pdf. After that, we use the Competitive Hebbian Learning to analyze the connectivity between the detected maxima of the pdf upon Mahalanobis distance. The so detected groups of Maxima are then used for the segmentation. Compared to the K-means clustering or to the clustering approaches based on the different competitive learning schemes, the proposed approach has proven, under a real and synthetic test images, that does not pass by any thresholding and does not require any prior information on the number of classes nor on the structure of their distributions in the dataset.

References

  1. Delaunay, B. (1934). Sur la Sphère vide. Bulletin of the Academy of Sciences USSR,VII, pp.793-800.
  2. Devijver, P.A., & Kittler, K. (1982) 'Pattern recognition: A statistical approach', Englewood Cliff, NJ, PrenticeHall international.
  3. Eddarouich, S., & Sbihi, A. (2007) 'Neural Network for Modes Detection in Pattern Classification'. ICTIS'07, Morocco, Fez, 3-5 pp. 300-303.
  4. Fritzke, B. (1995) 78 A growing neural gas network learns topologies', In G. Tesauro, D. S. Touretzky, & T. K. Leen (Eds.), Advances in neural information processing systems: 7. Cambridge, MA: MIT Press, pp. 625_632.
  5. Fukunagal, K., & Hostler, D. (1975). The estimation of the gradient of a density function with applications in pattern recognition. IEEE Trans. Inf. Theory, vol. IT21, n°1, p 32-40.
  6. Gray, R.M. (1984) 'Vector Quantization', ASSP Magazine, IEEE (Vol 1, Issue 2 ), ISSN :0740-7467, 1984, pp. 4-29.
  7. Hammouche, K. Diaf, M. and Postaire, J.-G. (2005) 'A clustering method based on multidimensional texture analysis', Pattern Recognition, pp. 1-13.
  8. Hebb, D. (1949) 'Organisation of Behavior', Wiley, New York.
  9. Knuth, D. E. (1973) 'The art of computer programming', Volume III: Sorting and searching. Reading, MA: Addison-Wesley.
  10. Martinetz, T.M. (1993) 'Competitive Hebbian learning rule forms perfectly topology preserving maps', (ICANN), Gielen S. and Kappen B. (eds), Springer, Heidelberg, pp. 427_434.
  11. Martinetz, T., & schulten K. (1994). Topology Representing Networks. Neural Networks. Vol. 7, No 3, pp. 507-522.
  12. Mizoguchi, R., & Shimura, S. (1976). Nonparametric learning without a teacher based on mode estimation. IEEE Trans. Comput., C-25(11), pp.1109-1117.
  13. Moussa. A, Sbihi. A and Postaire. J-G, (2008) 'A Markov random field model for mode detection in cluster analysis'. Patt. Recog. Letters 29, pp. 1197-1207.
  14. Muthanna, A. H.,Touahni, R., Sbihi, A., & Messoussi, R. and Eddarouich, S. (2010) 'Détection des modes d'une distribution de données multidimensionnelles par réseau de neurones et morphologie mathématique', Journées d'optique et du traitement de l'information.
  15. Parzen, E. (1962) 'An Estimation of a Probability Density Function and Mode', Ann. Math. Stat., vol. 33, pp. 1065-1076.
  16. Postaire, J.-G., & Vasseur, C. P. A. (1982) 'A fast Algorithm for non Parametric Probability Density Estimation', IEEE, Trans. on Pattern Anal. and Machine Intel. PAMI-4, n°6, pp. 663-666.
  17. Timouyas, M., Eddarouich, S., & Hammouch, A. (2012). A new approach of classification for non-Gaussian distribution upon competitive training, (ICCS'12), Agadir, Morocco, pp.1-6.
  18. Timouyas, M., Eddarouich, S., Hammouch, A, Touahni, R. & Sbihi, A. (2014)'Unsupervised NeuralMorphological Colour Image Segmentation Using the Mahalanobis as Criteria of Resemblance', ICMCS'14, Marrakech, Morocco, pp.314-320.
  19. Uchiyama, T., & Arbib, M. A. (1994) 'Color image segmentation using competitive learning', IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 16, n 12, pp. 1197-1206.
  20. Vasseur, C. P. A. & Postaire, J-G. (1980) 'A connexity testing method for cluster analysis' I.EEE Trans. Syst. Man. Cyber, vol SMC-10, n°3, p 145-179.
  21. Verikas, A., Malmqvist, K., & Bergman, L. (1997). Colour image segmentation by modular neural network. Pattern Recognition Letters, Vol 18, Issue 2, pp. 173-185.
  22. Yeo, N.C., Lee, K.H., Venkatesh, Y.V. and Ong, S.H. (2005) 'Colour image segmentation using the selforganizing map and adaptive resonance theory', Image and Vision Computing 23, pp. 1006-1079.
Download


Paper Citation


in Harvard Style

Timouyas M., Eddarouich S. and Hammouch A. (2016). Color Image Segmentation upon a New Unsupervised Approach using Amended Competitive Hebbian Learning . In Proceedings of the 18th International Conference on Enterprise Information Systems - Volume 2: ICEIS, ISBN 978-989-758-187-8, pages 205-210. DOI: 10.5220/0005918102050210


in Bibtex Style

@conference{iceis16,
author={Meriem Timouyas and Souad Eddarouich and Ahmed Hammouch},
title={Color Image Segmentation upon a New Unsupervised Approach using Amended Competitive Hebbian Learning},
booktitle={Proceedings of the 18th International Conference on Enterprise Information Systems - Volume 2: ICEIS,},
year={2016},
pages={205-210},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0005918102050210},
isbn={978-989-758-187-8},
}


in EndNote Style

TY - CONF
JO - Proceedings of the 18th International Conference on Enterprise Information Systems - Volume 2: ICEIS,
TI - Color Image Segmentation upon a New Unsupervised Approach using Amended Competitive Hebbian Learning
SN - 978-989-758-187-8
AU - Timouyas M.
AU - Eddarouich S.
AU - Hammouch A.
PY - 2016
SP - 205
EP - 210
DO - 10.5220/0005918102050210