Image segmentation with mean shift and region merging methods
Chunhui Zhao · Harbin Gongcheng Daxue Xuebao/Journal of Harbin Engineering University · 2008
In multi-cell segmentation process of medical images,segmentation quality declines when cell edges are blurred.To solve segmentation problem in such complex situations,this paper presented a combined segmentation method.This new method combined a nonparametric clustering method(mean shift) with region merging.First,an over-segmentation result was obtained by the mean shift algorithm.Then,based on the rules of region merging,the correct result was achieved by merging the over-segmentation regions.This method was applied to microscopy images of live cells.Its performance was proven more efficient than the geometric active contour algorithm,threshold segmentation algorithm,mean shift algorithm and watershed method.Using this method,overlapped cells can be individually separated,with the edges of individual cell in image approaching its real edges.This method lays a good foundation for further image research.