The Improved Two-dimensional Empirical Mode Decomposition And Its Application On Image Edge Detection

Huang proposed the empirical mode decomposition method in1998.EMD is a data-driven and adaptive method of signal processing, and appliesto deal with nonlinear non-stationary signals. French scholar J.Nunespromoted the idea of one-dimensional EMD, and put forward atwo-dimensional empirical mode decomposition method.This paper analyzes the decomposition of EMD and BEMD, andintroduces the method of extracting extreme points and the interpolationmethod of seeking envelope surface, and discusess sifting stop condition. Dueto the complexity of the empirical mode decomposition, there are still someproblems in practical applications, such as boundary effects, the selection ofscreening stop condition and so on. For solving these problems, an improvedtwo-dimensional empirical mode decomposition method is proposed. It usesthe method of extending the boundary data to suppress boundary effect. Itadds a part of data on the edge of the image, these data are the outwardtranslation of the edge data. This method does subsequent processing of theIMF components extracted to solve the problem of excessive screening.The improved BEMD is applied to the image edge detection in this thesis.Firstly, the source image is decomposed by BEMD; Secondly, we make thefirst IMF component with more edge information as the original image ofedge detection; finally, we do binarizing processing and morphologicalthinning for IMF. So the edges of the image are obtained. Experiments showthat the improved BEMD has achieved good results in simple image edgedetection.