基于分层策略的矿石颗粒图像区域分割  

Ore particle image region segmentation based on multilevel strategy

作　　者：杨广强[1] 王焕钢[1] 徐文立[1] 李鹏程[2] 王竹萌[2] 

机构地区：[1]清华大学自动化系,北京100084 [2]北京矿冶研究总院,北京100160

出　　处：《分析试验室》2014年第5期506-509,共4页Chinese Journal of Analysis Laboratory

基　　金：国家863计划资源环境技术领域主题项目(2011AA060203);国家科技支撑计划项目(2012BAB19B01)资助

摘　　要：为了消除过渡带对矿石颗粒图像多区域分割的不利影响,提出了一种基于分层策略的自动分割方法。该方法共分为3部分:首先,采用分层策略去除过渡带,构建图像金字塔;其次,基于多层灰度直方图定位平坦区域的特征峰,生成初始标记;最后,通过基于标记的区域生长完成图像分割。实际图像的分割结果验证了方法的有效性。与通用的多阈值方法相比,本方法能获得更为准确的分割结果。An automatic segmentation method based on multilevel strategy is proposed to eliminate the adverse impact caused by transitions on multi-region segmentation of ore particle images. The proposed method consists of 3 parts： firstly, transitions are eliminated through a multilevel strategy, meanwhile, an image pyramid is constructed; secondly, based on characteristic peaks of all flat regions acquired from histograms of multilevel images, initial marker is generated; thirdly, the segmentation is accomplished via the marker-based region growing method. Experiments on real images testify＂ the superiority of the proposed method over general multilevel threshold methods in the segmentation accuracy.

关 键 词：矿石颗粒图像 分层策略 图像分割 

分 类 号：O657.3[理学—分析化学]