Title: Colour image segmentation using spatial probabilistic clustering

Authors: Kirati Imène; Tlili Yamina

Addresses: Computer Science Department, Laboratory of Research in Computer Science (LRI), Badji Mokhtar-Annaba University, P.O. Box 12, 23000 Annaba, Algeria ' Computer Science Department, Laboratory of Research in Computer Science (LRI), Badji Mokhtar-Annaba University, P.O. Box 12, 23000 Annaba, Algeria

Abstract: This paper deals with colour image segmentation using spatial probabilistic clustering. The proposed method combines the robustness of non-parametric modelling with the spatial correlation of pixels. The probabilistic clustering handles the complex structures of the image without any assumption on the distributions of the image features. A summative function is used for weighting the class membership probabilities of pixels in order to favour the neighbouring pixels to belong to the same regions. Using the spatial information, the effect of noise is reduced and the obtained regions tend to be more homogenous. The tests on various images from the Berkeley image database demonstrate the efficiency of the proposed approach. Numerical results illustrate also the potential of the proposed approach compared to the state-of-the-art colour segmentation methods recently proposed in the literature.

Keywords: image segmentation; colour segmentation; probabilistic clustering; non-parametric modelling; spatial information; colour images; image processing.

DOI: 10.1504/IJSISE.2014.065265

International Journal of Signal and Imaging Systems Engineering, 2014 Vol.7 No.3, pp.173 - 179

Received: 28 Aug 2012
Accepted: 14 Aug 2013
Published online: 21 Oct 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article