结合SSFCM与随机游走的半监督图像分割算法  

Semi-supervised Image Segmentation based on Integration of SSFCM with Random Walks

作　　者：陈圣国[1,2] 孙正兴[1] 周杰[1] 李毅[1] 

机构地区：[1]南京大学计算机软件新技术国家重点实验室,南京210093 [2]金陵科技学院信息技术学院,南京211169

出　　处：《计算机辅助设计与图形学学报》2013年第7期1074-1082,共9页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61272219;61100110;61021062);国家"八六三"高技术研究发展计划(2007AA01Z334);江苏省科技支持计划(BY2012190;BE2011058;2010072)

摘　　要：针对基于颜色特征空间的半监督聚类分割算法适合分割结果包含多个颜色特征相似目标的应用场合,但对高噪声图像却无法获得理想的分割结果,而基于随机游走理论的半监督图像分割算法需要用户对目标逐一进行标记的问题,提出一种半监督图像分割算法.首先根据用户标记采用半监督模糊C均值聚类(SSFCM)算法对图像颜色特征进行建模;然后引入一个确信度函数,并根据SSFCM算法得到的隶属度数据计算确信度函数值,再将像素分为2类,分别作为随机游走图像分割算法的已标记点和未标记点;最后采用随机游走算法完成最终的分割.实验结果表明,该算法对图像中的噪声具有良好的抑制作用,且无需用户对目标逐一进行标记.Algorithms based on semi-supervised clustering are suitable to segment images containing a large amount of objects with the similar color features, but they cannot gain ideal effects to images containing noises; semi-supervised image segmentation algorithm based on random walks theory requires the user to label all objects contained in the image. A semi-supervised image segmentation algorithm to solve this problem is presented, which is based on integration of semi-supervised fuzzy c- means clustering algorithms with random walks. It models the image＇s color feature through semi- supervised c-means clustering algorithm（SSFCM） based label data, then it defines a reliability function based on the membership calculated by SSFCM, and the pixels are classified into two types that are considered as labeled and unlabeled pixels of Random Walks. The experimental results indicate that the algorithm not only reduces the noise sensitivity of SSFCM but also avoids cumbersome operations that the user labels the seed points of all objects for Random Walks.

关 键 词：半监督图像分割 半监督模糊C均值聚类 随机游走 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术]