Improving the unsupervised LBG clustering algorithm performance in image segmentation using principal component analysis

In this paper, a new method for improving unsupervised LBG clustering algorithm has been proposed. This algorithm belongs to the hard and K\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$K$$\end{document}-means vector quantization groups and drive directly from a simpler LBG. The defect of the LBG algorithm is to partition cluster in different iterations blindly. The basic idea of this paper is to use of principal component analysis and eigenvalue for handling this issue. Utilizing the eigenvalue in each step of LBG algorithm, it can either prevent from blindly splitting of vector or aggregation of data points in each cluster undoubtedly. The proficiency of eigenvalue-based LBG (E-LBG) algorithm is tested against other clustering algorithm such as Fuzzy c\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$c$$\end{document}-Means and Gustafson–Kessel. On standard database (Iris database) and acceptable results are obtained. Comparing the obtained result of simple LBG with E-LBG in term of time and accuracy has shown that the better performance of E-LBG method in segmentation of images.