摘要
为克服模糊C均值聚类图像分割对噪声较为敏感的缺陷,提出了一种结合协方差描述子的模糊C均值聚类算法。采用协方差描述子的滤波能力以改善传统模糊C均值聚类算法对噪声敏感的缺陷;提取超像素的协方差矩阵作为特征,降低图像识别的特征冗余。并做了仿真实验,对提出的算法与三个图像分割算法进行比较,结果表明该图像分割算法具有较好的噪声鲁棒性和分割准确率。
This paper presents a new method to overcome the drawback of fuzzy c-means cluster in noiserobust. It proposes to apply covariance descriptor to superpixel as a feature. Then it introduce the FCMbased algorithm incorporating covariance descriptor. The experiments reveals that the approach achieves competitive and even better results compared with three FCM-based algorithms.
出处
《信息技术》
2016年第3期97-100,共4页
Information Technology
基金
南京领军型科技创业人才引进计划(2014A090002)
关键词
图像分割
模糊C均值聚类
超像素
协方差描述子
image segmentation
fuzzy C-means cluster
superpixel
covariance descriptor
