Abstract:An interactive image segmentation algorithm based on Mean-Shift cluster was proposed. Firstly, image was segmented coarsely by Mean-Shift cluster algorithm. Secondly, users only need to input interactive information simply and indicate the location of object points and background points roughly. Finally, maximal similarity principle based on region merging was used to segment region of interest. Experimental results show this method is simple and effective, and can extract object counter from complex background exactly.