结合均值漂移与最小生成树的图像分割算法

为了提高均值漂移(MS)分割算法的运行效率,提出了一种结合MS与最小生成树(MST)的图像分割方法,简称MS-MST方法。首先选取较小的空间带宽参数,以较快的速度对图像进行MS分割,得到过分割图像;然后,以过分割区域作为后续处理的基本单元,构造加权区域邻接图,运用MST算法对其进行合并,得到最终的分割结果。实验结果表明,本文算法在保证图像分割质量的前提下,大幅提高了经典MS算法的分割速度。

Image segmentation algorithm combining mean shift with minimum spanning tree

This paper presents an image segmentation algorithm combining mean-shift with the minimum spanning tree,in order to improve the operating efficiency of the classic mean shift.The algorithm first selects a smaller spatial bandwidth,and applies the mean shift to over-segment image at a faster speed.Then,we regard the over-segmentation region as the basic unit of subsequent procedure to construct a weighted region adjacency graph,and then use the minimum spanning tree algorithm to merge over-segmentation image.The experimental results verify that this algorithm,on the premise of ensuring the quality of image segmentation,substantially increases the speed of the classic mean shift segmentation algorithm.