Automatic centroids selection in K-means clustering based image segmentation

This paper proposes a K-means clustering based image segmentation algorithm which select the centroids automatically. It eliminates the limitations associated with K-means clustering such as selection of initial centroids and dead centers. As image histogram is one of the best ways to represent the distribution of image gray levels, the proposed approach selects centroids as the gray levels corresponding to the peaks of the image histogram. With these initial centroids, K-means clustering is performed. The result is post processed by some morphological operations. The proposed algorithm uniformly segments the regions of interest over randomly selected centroids. The performance of proposed algorithm and random centroids selection is compared with some validity parameters like SSIM, MSE, PSNR, IF, SC and CC. Comparison with the existing algorithms confirms the improvement in qualitative parameters. The tool used in this work is MATLAB R2012a.

[1]  Anil K. Jain,et al.  Data clustering: a review , 1999, CSUR.

[2]  T. Thomas,et al.  Clustering of Invariance Improved Legendre Moment Descriptor for Content Based Image Retrieval , 2008, 2008 International Conference on Signal Processing, Communications and Networking.

[3]  Ramachandran Baskaran,et al.  A Survey on Internal Validity Measure for Cluster Validation , 2010 .

[4]  M. P. S Bhatia,et al.  Data clustering with modified K-means algorithm , 2011, 2011 International Conference on Recent Trends in Information Technology (ICRTIT).

[5]  Daoliang Li,et al.  An improved KK-means clustering algorithm for fish image segmentation , 2013, Math. Comput. Model..

[6]  Wilhelm Burger,et al.  Digital Image Processing - An Algorithmic Introduction using Java , 2008, Texts in Computer Science.

[7]  M.P. Sebastian,et al.  Enhancing the K-means Clustering Algorithm by Using a O(n logn) Heuristic Method for Finding Better Initial Centroids , 2011, 2011 Second International Conference on Emerging Applications of Information Technology.

[8]  N. Otsu.  A threshold selection method from gray level histograms , 1979 .

[9]  Nasser M. Nasrabadi,et al.  Image coding using vector quantization: a review , 1988, IEEE Trans. Commun..

[10]  Rui Xu,et al.  Survey of clustering algorithms , 2005, IEEE Transactions on Neural Networks.

[11]  Punya Thitimajshima,et al.  A new modified fuzzy c-means algorithm for multispectral satellite images segmentation , 2000, IGARSS 2000. IEEE 2000 International Geoscience and Remote Sensing Symposium. Taking the Pulse of the Planet: The Role of Remote Sensing in Managing the Environment. Proceedings (Cat. No.00CH37120).

[12]  Eero P. Simoncelli,et al.  Image quality assessment: from error visibility to structural similarity , 2004, IEEE Transactions on Image Processing.

[13]  M. M. Rahman,et al.  Improvement of K-means clustering algorithm with better initial centroids based on weighted average , 2012, 2012 7th International Conference on Electrical and Computer Engineering.

[14]  David M. Mount,et al.  A Fast Implementation of the Isodata Clustering Algorithm , 2007, Int. J. Comput. Geom. Appl..

[15]  J. MacQueen.  Some methods for classification and analysis of multivariate observations , 1967 .