A Boundary Detection algorithm of clusters based on Dual Threshold Segmentation
Baozhi Qiu, Shuang Wang · 2011
Boundary points detection of clusters is important in image processing, machine learning and so on. We propose a boundary detecting algorithm called BDDTS (Boundary Detection algorithm of clusters based on Dual Threshold Segmentation), which is based on the characteristics of the distribution of boundary points. The algorithm, which firstly accrues to the different cost function values of data points, then divides data set into the internal point set, the intermediate point set and the external point set. Secondly by removing the internal points from intermediate point set and combining it to external point set, we get candidate boundary set. At last, we exploit the secondary processing for the candidate boundary set in order to obtain a more accurate boundary. The experimental results show that BDDTS can detect boundary points of clusters in arbitrary shapes, size and densities very rapidly and efficiently. It is also applicable to real data set and high-dimensional data set.