基于SOM-K聚类的图像分割算法  

An image segmentation algorithm based on SOM-K clustering

作　　者：张亚楠 杜宇人[1] 

机构地区：[1]扬州大学信息工程学院,江苏扬州225127

出　　处：《扬州大学学报（自然科学版）》2017年第3期49-52,共4页Journal of Yangzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(51273172)

摘　　要：针对介电弹性体图像在电致形变实验中存在噪声过多、光照不均和无法准确分割等缺陷,提出一种基于形态学重建、核聚类算法与分水岭算法相结合的目标图像分割改进算法.首先选取合适的结构元素对图像进行形态学重建,然后对重建后的形态学梯度图像进行分水岭分割,最后通过SOM-K算法聚类所得分割图像,合并相似区域,以降低过分割的影响.结果表明该分割算法准确度高且可行性好.According to the disadvantages of noise,uneven illumination and inaccurate segmentation in the images of dielectric elastomer,the author creates a new method based on morphological reconstruction,kernel clustering algorithm and watershed algorithm.This method selects the appropriate structural elements for morphological reconstruction of the image first of all and then segments image by watershed algorithm.Finally,it clusters the image segmented based on SOM-K clustering algorithm and merges the similar regions to reduce the influence of over segmentation.The experiments show that the proposed method not only segments accurately in a real time,but also works better in the indexes of shape measure,uniformity measure and gray-level contrast the other methods.

关 键 词：图像分割 介电弹性体 分水岭算法 形态学重建 核聚类算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]