doi:10.1016/j.patrec.2007.03.012 
Copyright © 2007 Elsevier B.V. All rights reserved.
Image segmentation by clustering of spatial patterns
Yong Xiaa, b, 
, 
, (David) Dagan Fenga, c, Tianjiao Wangd, Rongchun Zhaob and Yanning Zhangb
aSchool of Information Technologies, J12, The University of Sydney, Sydney, NSW 2006, Australia
bSchool of Computer Science, Northwestern Polytechnical University, Xi’an 710072, China
cCenter for Multimedia Signal Processing, Department of Electronic and Information Engineering, Hong Kong Polytechnic University, Hong Kong
dSchool of Automation, Northwestern Polytechnical University, Xi’an 710072, China
Received 11 February 2006;
revised 8 January 2007.
Communicated by Y.J. Zhang.
Available online 31 March 2007.
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Abstract
This letter describes an approach to perceptual segmentation of images through the means of clustering of spatial patterns. An image is modeled as a set of spatial patterns defined on a rectangular lattice. The distance between a spatial pattern and each cluster is defined as a combination of the Euclidean distance in the feature space and the spatial dissimilarity which reflects how much of the pattern’s neighbourhood is occupied by other clusters. Our approach has been compared with the Fuzzy C-Mean (FCM) algorithm, a spatial fuzzy clustering algorithm and a Markov Random Field (MRF) based algorithm by segmenting synthetic images, texture mosaics and natural images. The results of those comparative experiments demonstrate that the proposed approach can segment images more effectively and provide more robust segmentation results.
Keywords: Image segmentation; Image texture analysis; Spatial pattern; Fuzzy clustering
Fig. 1. An example plot of the weighting coefficient α(n) vs. the iteration number n.
Fig. 2. Two test cases of synthetic image (σ = 0.015 and 0.065) and their segmentations by applying (the 2nd column) the FCM algorithm, (the 3rd column) the SCF algorithm, (the 4th column) the MRF algorithm and (right column) the proposed algorithm.
Fig. 3. Plot of error percentage of incorrectly classified pixels on image set GIII.
Fig. 4. Four test cases of mosaics of four textures (MIV1–MIV4) and their segmentations by applying (the 2nd column) the SCF algorithm, (the 3rd column) the MRF algorithm and (right column) the proposed algorithm.
Fig. 5. Two real scene images and their segmentations by applying (the 2nd column) the SCF algorithm, (the 3rd column) the MRF algorithm and (right column) the proposed algorithm.
Table 1.
Twelve natural textures from Brodatz album
Table 2.
Error percentage of incorrectly classified pixels on image set MIV
Table 3.
Time cost of three segmentation algorithms on image set MIV
Abstract
Purchase PDF (982 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (3)
Purchase PDF (588 K)
