A fusion algorithm for remote sensing images based on nonsubsampled pyramids and bidimensional empirical decomposition  

作　　者：ZHANG XiaoDong 1,WANG WenBo 2,WANG DiFeng 3 & ZHANG Yu 4 1 State Key Laboratory of Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,Wuhan 430079,China 2 College of Sciences,Wuhan University of Science and Technology,Wuhan 430065,China 3State Key Laboratory of Satellite Ocean Environment Dynamics,Second Institute of Oceanography,State Oceanic Administration,Hangzhou 310012,China 4 Spatial Information Technology Application Research Institute,Changjiang River Scientific Research Institute,Wuhan 430010,China 

出　　处：《Science China(Technological Sciences)》2010年第S1期196-204,共9页中国科学（技术科学英文版）

基　　金：supported by the National Basic Research Program ofChina("973"Program)(Grant Nos.2006CB701300,2006CB701304);the China Postdoctoral Foundation(Grant No.2007041172),Hubei Natural Science Foundation(Grant No.2007ABA042);LIESMARS Special Research Fund and the Wuhan Key Scientific and Technological Project(Grant No.200810321144)

摘　　要：In order to improve the quality of remote sensing image fusion,a new method combining nonsubsampled Laplacian pyramid (NLP)and bidimensional empirical mode decomposition(BEMD)is proposed.First,the high resolution panchromatic image (PAN)is decomposed using NLP until the approximate component and the low resolution multispectral image(MS)contain features with a similar scale.Then,the approximation component and the MS are decomposed by BEMD,resulting in a number of bidimensional intrinsic mode functions(BIMF)and a residue respectively.The instantaneous frequency is computed in 4 directions of the BIMFs.Considering the positive or negative coefficients in the corresponding position,a weighted algorithm is designed for fusing the high frequency details using the instantaneous frequency and the coefficient absolute value of the BIMFs as fusion feature.The fused image is then obtained through inverse BEMD and NLP.Experimental results have illustrated the advantage of this method over the IHS,DWT andà-Trous wavelet in both spectral and spatial detail qualities.In order to improve the quality of remote sensing image fusion,a new method combining nonsubsampled Laplacian pyramid (NLP)and bidimensional empirical mode decomposition(BEMD)is proposed.First,the high resolution panchromatic image (PAN)is decomposed using NLP until the approximate component and the low resolution multispectral image(MS)contain features with a similar scale.Then,the approximation component and the MS are decomposed by BEMD,resulting in a number of bidimensional intrinsic mode functions(BIMF)and a residue respectively.The instantaneous frequency is computed in 4 directions of the BIMFs.Considering the positive or negative coefficients in the corresponding position,a weighted algorithm is designed for fusing the high frequency details using the instantaneous frequency and the coefficient absolute value of the BIMFs as fusion feature.The fused image is then obtained through inverse BEMD and NLP.Experimental results have illustrated the advantage of this method over the IHS,DWT andà-Trous wavelet in both spectral and spatial detail qualities.

关 键 词：bidimensional empirical mode DECOMPOSITION nonsubsampled PYRAMID instantaneous frequency REMOTE sensing image FUSION 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]