摘要
传统模糊C均值算法没有充分利用像素周围的空间信息,所以算法抗噪效果不理想,且该算法仅利用像素隶属度信息,分割规则过于单一。因此,提出一种基于包含度及空间信息的聚类算法以提高图像分割抗噪性和准确性。首先将包含度信息加入到目标函数中,以弥补隶属度单一化的不足;其次将像素周围的邻域信息作为空间信息加入到目标函数中,使用信息熵与交叉熵调节像素信息和空间信息之间的权重;最后使用梯度下降法优化该目标函数以便对图像进行正确分割。以4组卫星图像为例进行分割,并分别与FCM算法、PCM算法、AFCM_S1算法进行对比。实验结果表明,基于包含度和空间信息的聚类算法对噪声点具有较好的处理效果,可提升分割精度和负率度。
Since fuzzy C means algorithm does not make full use the spatial information,it is not ideal for noise reduction,and this algorithm applies membership degree as the only segmentation criterion.To solve these problems,this article proposes a novel clustering model based on inclusion degree and spatial information for enhancing noise resistance and accuracy of image segmentation.Firstly,this model adds inclusion degree into objective function for making up the shortage of membership degree.Secondly,this model adds spatial information into objective function and applies information entropy and cross entropy to obtain the weights of pixel information and spatial information.Three sets of satellite images are used for showing the goodness of the proposed method,and it is also compared with the FCM algorithm and AFCM_S1 algorithm respectively.According to the experimental results,this novel model has a good effect on noise resistance and improves the ACC and NRM.
作者
李咨兴
唐坚刚
刘丛
LI Zi-xing;TANG Jian-gang;LIU Cong(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《软件导刊》
2019年第2期148-152,共5页
Software Guide
关键词
模糊C均值
聚类分析
图像分割
空间信息
包含度
fuzzy C-mean
clustering analysis
image segmentation
spatial information
inclusion degree
