A Speckle Reduction Algorithm for SAR Image Based on Complex Wavelet Domain
Xiao Ke Zhu · Signal Processing · 2009
Synthetic aperture radar (SAR) images are inherently affected by multiplicative speckle noise.The despeckling process should suppress the noise as possible on the condition that the detail information is well retained.In this paper,a novel frame- work using Ganssian scale mixtures (GSM) model based on complex wavelets domain is proposed.Neighborhoods of coefficients at adja- cent scales in dual-tree complex wavelet transform (DTCWT) are constructed as the product of a Ganssian variable and a hidden multi- plier.Under this model,a Bayes Least Squares (BLS) estimator is adopted to reduce speckle noise.According to the results which are e- valuated on simulated and real SAR image experiments,the proposed algorithm performs more effectively than traditional spatial filters and wavelet methods in terms of both speckle reduction and signal detail preservation,the approving visual effect is thus obtained.