基于Hausdorff测度的图像分割算法研究  

Study on Image Segmentation Algorithm Based on Hausdorff Measure

作　　者：廖定安[1] 肖海慧[1] 王成杰[1] LIAO Ding-an;XIAO Hai-hui;WANG Cheng-jie(Changzhou Vocational Institute of Textile and Garment,Changzhou 213164 China)

机构地区：[1]常州纺织服装职业技术学院

出　　处：《自动化技术与应用》2019年第8期89-92,共4页Techniques of Automation and Applications

基　　金：江苏省第五期“333工程”科研资助项目(编号BRA2017484);常州纺织服装职业技术学院重点课题(编号51800011452)

摘　　要：在对弱边界的图像进行分割时,边界区域的像素点分类问题是一个不确定性问题,对于此类问题,无论是随机过程的分布函数,还是模糊数学的隶属函数往往都不易确定。传统基于欧氏距离测度对像素点归类的方法仅利用了像素的灰度信息,而忽视空间位置信息。提出一种加权的Hausdorff距离测度,利用像素邻域区间分析方法,力图更加接近复杂不确定性问题的求解,将该测度应用于传统FCM、KFCM算法,对Brainweb脑部图像进行实验,结果表明,应用加权的Hausdorff距离测度方法比传统欧氏距离测度方法分类正确率更高。In the segmentation of the weak boundary image,the pixel classification problem of the boundary region is an uncertain problem.For such problems,the distribution function of the random process and the membership function of the fuzzy mathematics are often not easy to determine.The traditional method of classifying pixels based on Euclidean distance measure takes advantage of pixel gray information instead of spatial location information.A weighted Hausdorff distance measure is proposed,and the method of pixel neighborhood interval analysis is used to get closer to the solution of the complex uncertainty.The measure is applied to the traditional FCM and KFCM algorithm,and the Brainweb brain image are tested.The results show that we get higher accuracy with the weighted Hausdorff Distance measure methodthan with the traditional Euclidean distance measure method.

关 键 词：图像分割 加权Hausdorff 弱边界 欧氏测度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]