A Non-parametric Image Segmentation Algorithm Using an Orthogonal Experimental Design Based Hill-Climbing

Image segmentation is an important process in image processing. Clustering-based image segmentation algorithms have a number of advantages such as continuous contour and non-threshold. However, most of the clustering-based image segmentation algorithms may occur an oversegmentation problem or need numerous control parameters depending on image. In this paper, a non-parametric clustering-based image segmentation algorithm using an orthogonal experimental design based hill-climbing is proposed. For solving the oversegmentation problem, a general-purpose evaluation function is used in the algorithm. Experimental results of natural images demonstrate the effectiveness of the proposed algorithm.