摘要
为衰减存在于图像上的噪声,尽可能地精确提取目标,提出一种基于样本和属性加权的二维模糊C-均值(FCM)聚类分割法。构造合理的二维直方图对图像进行滤波,通过改变样本权的幂函数大小提高目标提取的精确性,为提高聚类效果及抗噪性,对目标函数的类内距离进行属性加权,实现属性权数值的自动确定。基于遥感图像和小目标图像的实验结果表明了该方法的有效性。
To improve the anti-noise property and extract target image efficiently,sample and attribute weighting fuzzy C-means clustering（FCM）algorithm based on two-dimensional histogram（2DH）was proposed.A reasonable 2DH was created to filter the original image.Power function of sample weight was changed to improve the accuracy of target extraction.For improving the clustering effect and anti-noise property,attribute weight was applied to inner-class distance of target function,and the values of attribute weight was determined automatically.Experimental results based on the remote sensing and small target images illustrate the effectiveness of the proposed method.
出处
《计算机工程与设计》
北大核心
2016年第6期1604-1609,1631,共7页
Computer Engineering and Design
基金
国家自然科学基金重点项目(61136002)
陕西省自然科学基金项目(2014JM8331
2014JQ5138)
国家863高技术研究发展计划基金项目(2013AA014504)
陕西省教育厅科学研究计划基金项目(2015JK1654)
关键词
模糊C-均值聚类
二维直方图
样本加权
属性加权
类内距离
fuzzy C-means clustering
two-dimensional histogram
sample weight
attribute weight
inner-class distance
