一种快速的二维经验模式分解算法  

A Fast Bidimensional Empirical Mode Decomposition Algorithm

作　　者：秦绪佳[1,2] 范颖琳[1] 郑琴[1] 郑红波[1,2] 徐晓刚[3] 

机构地区：[1]浙江工业大学计算机科学与技术学院,杭州310032 [2]浙江省可视媒体智能处理技术研究重点实验室,杭州310032 [3]大连舰艇学院装备系统与自动化系,辽宁大连116018

出　　处：《小型微型计算机系统》2013年第9期2182-2187,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61075118;61273262)资助;浙江省自然科学基金项目(Y1100880)资助

摘　　要：作为分析非线性、非平稳信号的有效方法,经验模式分解将原始信号解析成不同时间尺度下的平稳的数据分量及趋势项集合.将其应用到图像处理领域,图像可以被分解成一系列的图像细节信息及趋势信息.这种层次式的图像表示形式,使得图像在融合、边缘检测、滤波及纹理分析等方面产生更优的结果.提出一种邻域限制顺序统计滤波器的快速自适应二维经验模式分解算法.算法在时域内对每次分解的最大邻域进行了限制,首先在最大邻域内计算出局部极大值点集和局部极小值点集,然后对局部极值点集采用基于顺序统计滤波器的包络估计算法来计算信号的上下包络.实验结果证明,本文算法具有较快的分解速度,有效的防止了灰度斑在分解结果中出现,同时具有较强的信号细节保持能力.As an effective way of analyzing non-linear and non-stationary signals, empirical mode decomposition makes the original signal be analyzed into a set of data layers with different time-scales and the trend. While it is applied in the area of image processing, images can he decomposed into a series of detail information and some descriptions about the global trends. This model of image pres- entation makes the result of some image processing much more high-quality, such as image fusion, edge detection, image filtering and texture analysis. In this paper, a fast and adaptive bidimensional empirical mode decomposition algorithm based on neighborhood limited and order-statistics filter is proposed. This algorithm sets a max neighborhood for each decomposition in time domain and in this max neighborhood calculates local maximum point sets and local minimum point sets , then by those extreme point sets calculates the upper and lower envelope of signal using envelope estimation algorithm based on order-statistics filter. Experimental results show that this algorithm has a faster decomposition speed and effectively prevents the gray spots in the decomposition results, while it has stronger ability of maintaining the details of the signal.

关 键 词：经验模式分解 限邻域 顺序统计滤波器 自适应 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]