Research On The Key Problems Of Bidimensional Empirical Mode Decomposition In Digital Image Processing

Bi-dimensional Empirical Mode Decomposition(BEMD)is a technique extended in bi-dimensional case from the Huang transform of empirical mode decomposition.This technique provides a new approach to meet the requirement of analyzing non-stationary signal features in image processing and actual needs of mining enterprises,due to its good data-driven adaptive ability.However,in its applications,it has been found some problems associated with the interpolation optimization,end effects,stopping criteria,and disaggregation of the decomposed images in image fusion.To solve these problems,this dissertation undertakes a systematic investigation for four aspects of contents in detail as follows.(1)An interpolation technique based on the particle swarm optimization and the fractal description is proposed.By means of the fractal Brownian function the feature quantities of an image are obtained,and then the interpolation is operated.The technique not only improves the performance of interpolation with so high accuracy and efficiency that a quicker run of the BEMD can be seen,but also lays the good foundation of solving several other problems.(2)An approach to treatment of the end effect is presented by means of the combination of the adaptive support vector extension with the mirror closure technique.The adaptive support vector extension results in an extrapolation image from which all extreme points closest to and outside the boundary are picked up,and the mirror closure technique is used to form a closed image by the following mirror extension operation.Results from the related images show that the proposed approach weakens or even eliminates end effects of the BEMD,which may ensure reliability of the information extracting from edge parts and details of the image.(3)A stopping criterion is established on the basis of the number of extreme points and its variation rate when the sieve surfaces give the projection of extreme points at different locations on the zero-mean-value plane.At first,to analyze the envelope surfaces obtained in the process of the BEMD,to trace the evolution of the extreme points in the whole process of decomposition,and then to form a judgment whether the screening stops or not on the position information of the continuous screening surfaces.In this way,the bi-dimensional intrinsic mode function(BIMF)obtained by the BEMD may represent well features of the image itself,and the phenomena of over and under decomposition are effectively eliminated.(4)Two new algorithms,the image feature extraction of BEMD-SIFT and the image fusion of BEMD-Extreme point coordination,are established.The former is based on the synthesis of summing those components of BIMFs yielded from the BEMD in terms of the extracted features for all of them.The algorithm,free from the mutual interference between different modes in information,is beneficial to more quickly and more accurately obtain various features of the image.The latter introduces a coordination operation of multiple images,which is equivalent to find the adaptive basis function of the images under consideration by means of a set of maximum and minimum points yielded from the BEMD.In the frame of an identical adaptive basis function,all of the BIMFs of the decomposition are matched together,and are fused and synthesized into the cumulative fusion image,which leads to the image fusion algorithm with adaptive ability.Applicaion results show that the presented algorithms can preserve information details and edge features of the original images more satisfactorily in the multi-scale image fusion,which owns great advantages for improving quality of the fusion image.A large number of illustrated implementation cases in this dissertation verify good application effects of the proposed technical measures in image processing.This work suggests a systematic research on the development of the BEMD technique and the improvement of its application to image processing.It also provides some beneficial fundamentals and technical tools for analyzing of non-stationary signal characteristics in the field of image processing.At the same time,it is also the application of adaptive digital image processing in mine safety monitoring.