Abstract:

By combining the empirical mode decomposition(EMD) and principle component analysis(PCA), a new speckle reduction method based EMD is proposed for logarithmic SAR images. Based on the Statistical properties of logarithmically transformed speckle and energy distribution model of decomposed Gaussian white noise by EMD, the noise energy in each intrinsic mode function is approximately calculated. After the intrinsic mode function is decomposed by PCA, to further remove the noise and keep useful details, we select the appropriate principal components according to the noise energy proportion in intrinsic mode function and the decomposition characteristics of PCA, then reconstruct the intrinsic mode function by the selected principal components. The experimental results show that the SAR image edges are retained better and the speckles are removed effectively with the proposed method.

Key words: empirical mode decomposition, SAR image, speckle reduction, principle component analysis