Research On Synthetic Aperture Radar Image Despeckling Algorithm Based On Heterogeneity Measure

As a new microwave remote sensing technology,Synthetic Aperture Radar(SAR)has been widely used in geological exploration,disaster monitoring,marine probe,environmental monitoring,military systems and so on.However,speckle inevitably exits in SAR images due to SAR's coherent imaging systems,which seriously degrades the image interpretation and scene analysis.Thus,Speckle suppression is a critical step in SAR image processing.An ideal despeckling algorithm should smooth the speckle and reserve image details at the same time.but none can satisfy these two demands perfectly.In recent years,heterogeneity measure and analysis has received increasing attention in the field of SAR image processing.The heterogeneity reflects the difference of regional information,by which the variation of texture information and the back scatterers can be distinguished.Taken heterogeneity measure as focus,this paper gives a deep investigation to SAR image despeckling in wavelet domain.The main contributions of this thesis are given below.(1)The representation and statistical distribution of the multiscale heterogeneity measure in SAR image are studied.A theoretical distribution is presented for this heterogeneity measure and the estimation of model parameter is given.Specifically,we adopt the multiscale local coefficient of variation(MLCV)as heterogeneity measure.Logarithmic normal distribution is proposed to model the histogram distribution and estimate the mode of MLCV.(2)A bayesian wavelet speckle reduction algorithm based on heterogeneity classification is developed.Under the non-homomorphic framework,we use Normal inverse Gaussian(NIG)function for modeling backscattered signal in wavelet domain,and Gaussian function for speckle noise.The estimation formula of noise-free signal is derived by Bayesian maximum a posteriori(MAP)criterion.In order to improve the estimation precision of model parameters,we introduce multiscale local coefficient of variation(MLCV)as heterogeneity measure.Based on heterogeneity measure,each coefficient in wavelet sub-band is classified into one of several different heterogeneity scenes,and NIG model parameters are computed in each class through cumulants estimation method.Simulation experiments verify the feasibility and effectiveness of the proposed algorithm.(3)Over smoothing of image details is common in despeckling procedure for real SAR image.This thesis presents a method that first performs pre-correction on speckled image and then applies despeckling algorithm.Specifically,wavelet transform is used to decompose the noisy image,and multi-scale local coefficient of variation(MLCV)is adopted as heterogeneity measure.An adaptive pre-correction function is proposed based on the MLCV.Wavelet coefficients in each sub-band are classified into four categories,and different pre-processing strategies are taken into account correspondingly.Finally,a common wavelet domain despeckling algorithm is used to pre-correction image.Experiment on real SAR images shows that performance of the proposed method has been significantly improved both in speckle suppression and image details preservation.Due to simplicity and easiness to implement,the proposed pre-correction could be extended and combined with some conventional despeckling approaches,which is helpful in practical application.