BEMD方法的若干研究及改进  

Some Research and Improvement of BEMD Method

作　　者：苑宇[1,2] 李聪明[2] 赵玉龙[1] 

机构地区：[1]西安交通大学机械工程学院,西安710049 [2]大连交通大学动车运用与维护工程学院,大连116028

出　　处：《振动．测试与诊断》2013年第S1期69-72,220,共5页Journal of Vibration,Measurement & Diagnosis

基　　金：国家重点实验室开放课题研究基金资助项目(sklms2012006);辽宁省教育厅基金资助项目(L2012166)

摘　　要：对传统的二维经验模态方法进行研究。首先,对于边界效应主要通过对图像进行镜像延拓的方法减小边界对内部数据的污染;其次,观察相邻两次分解的图像极值点,比较在位置分布上的重合度和在数量增减速率上的变化情况,得出分量分解的收敛性规律;最后,结合大量分解实验与实际需求,得出满足实时性、准确性要求的分量筛选停止准则和分解过程停止准则的表达式。通过改进的优化方法,使传统方法得到一定程度的改善,也使二维经验模式分解(two-dimensional empirical mode decomposition,简称BEMD)理论方法得到完善,为图像处理以及其他二维信号的分析提供了精准、有效的分析方法。Compared with one-dimension EMD, there are still some disadvantages and shortcomings in bidimensional EMD method, including the smooth continuity of curved surface interpolation, the interior pollution of boundary effect, the sifting stop criteria of the component and the rationality of BEMD decomposition. To begin with, this paper researches deep into the traditional BEMD, starting with above problems. Firstly, to deal with the boundary effect, this paper takes the method of mirror image prolongation. Secondly, comparing the overlap ratio and the sum changing rate of the extreme points of two adjacent siftings can obtain astringency law. Last, according to the need of experiment and actual condition, this paper puts forward sifting stop criteria of the component and the stop expression of BEMD.To a certain extent, the improved optimization method can make the inadequacies in traditional BEMD get some improvements, and also makes the two-dimensional EMD theory method more perfect, and provides a more accurate, stable, and effective analysis method for improved image processing.

关 键 词：二维经验模式分解 边界效应 统计分析 停止准则 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]