基于形态学重建和极大值标记的分水岭分割算法

摘要

For over-segmentation problem of traditional watershed segmentation algorithm,this paper proposed a new segmentation algorithm,which combined morphology reconstruction and maxima mark.First,opening and closing recon-struction was conducted on the image,which could remove all kinds of noise and retain the original structure of the target. Secondly,local maxima was adj usted and marked,which could avoid the over-segmentation effectively.Finally,watershed segmentation on the local maxima image achieved good results.The key factor of the algorithm to select structural elements was discussed.A large number of experiments shows that selecting structural elements,whose shape is similar to the target image and whose size is appropriate to it,can achieve better segmentation.%针对传统分水岭算法的过分割问题，结合形态学重建和极大值标记技术研究并实现一种新算法。该算法首先进行开闭重建预处理，较好地去除了各类噪声并保留目标原有的结构；然后进行局部极大值调整及标记，有效地避免了过分割；最后对局部极大值图像进行分水岭分割，取得了较好的实验效果。论文还探讨该算法的关键因素即结构元素的选择问题，经大量实验得出结论———选择与目标图像形状相似、大小相宜的结构元素可获得较好的分割效果。