The Best Segmentation Of Images Based On The Gs Method

Clustering analysis has been employed in image segmentation for a long time. It is aimed at the best output of image segmentation in this paper, which focuses on the estimating the best number of clusters in a set of data by Gap Statistic (GS) Method.GS method, which is used of estimating the number of clusters in a set of data, was proposed by three scholars in Stanford University. In fact, cluster analysis just tells us how to cluster data, does not get a rule to judge the best cluster number. GS method proposes a Gap statistic by comparing the data dispersion with an average dispersion under an appropriate reference distribution, and to be used to estimate the best cluster number.GS method is the most important part of this paper. For applying it to image segmentation, a framework of image segmentation based upon the GS Method is proposed, which combined with clustering analysis. There are four part of it: image input, dealing with image feature, clustering and choosing the best and output the result.To get an operational procedure, the GS method has been modified, and the concretely expression of the Gap statistic is computed in the condition of 1-D by amending the Gap statistic. Furthermore, an algorithm is put forward to segment image by choosing threshold: the best adaptive k -threshold segmentation. As a matter of fact, the algorithm develops the framework—chose the best threshold to segment.Finally, the property of the algorithm has been analyzed base upon the result of enough image segmentations, and some amelioration has been advanced for the algorithm's deficiency.