Research On The Bi-dimensional Empirical Mode Decomposition And Application To The Analysis Of Synthetic Aperture Radar Image

Since the1990s Synthetic Aperture Radar (SAR) has been developed by leapsand bounds, as its imaging technology is becoming mature, the SAR image analysishas become an important research topic. However, the distribution in space of thepixel of SAR image is very complex, so SAR image is non-stationary and non-linearsignal, and it is difficult to research into analysis and application of SAR image. Someof the traditional methods such as Fourier analysis or wavelet analysis do not achievethe desired results in the SAR image processing. However, Empirical ModeDecomposition (EMD) has good local adaptability so that it has some advantages inprocessing non-stationary and non-linear signal. Consequently, it is necessary to studyBi-dimensional Empirical Mode Decomposition (BEMD) and to apply BEMD toSAR image analysis.This thesis analyzes the one-dimensional EMD algorithm and Hilbert transform,and studies the mathematical properties of EMD such as orthogonality, completenessand so on.The research focus of this article is study of BEMD algorithm and its applicationto SAR image analysis. On study of BEMD algorithm, the BEMD algorithmpreviously proposed has been improved. The improvements include selection ofcharacterized point, Plane Triangulation, choice of the judgment conditions ofintrinsic mode function(IMF) and choice of the program termination conditions. Inorder to make the SAR image processing more efficient, a fast BEMD algorithmbased on image segmentation and parallel processing is proposed.On SAR image denoising, BEMD algorithm is used to decompose SAR image,then several layers of IMF are denoised and superimposed upon the remaining layersignal, so the superimposed result is that the image has denoised. On SAR WaveSpectrum Inversion, the BEMD algorithm is used to decompose SAR image whose wave information is not obvious, then the several layers of IMF are processed by MPIwave spectral inversion algorithm to extract the wave spectrum. The experimentalresults show that the effect of the combination between the BEMD algorithm anddenoising and wave spectral inversion for SAR image has some increase comparedwith traditional methods, so this provides a new way for SAR image analysis.