A novel curvelet domain speckle suppression method for SAR images

This paper introduces a novel Bayesian method for speckle suppression of SAR images. We first analyze the logarithmic transform of the original image by means of the curvelet transform that handles image edges more efficiently than wavelet transform. In a recent work [1], we have shown that due to the statistical properties of the curvelet subbands of SAR images, they can be modelled by two-dimensional Generalized Autoregressive Conditional Heteroscedastic (2D-GARCH) model. Here, we employ a generalization of 2D-GARCH model, called 2D-GARCH Generalized Gaussian (2D-GARCH-GG), to these coefficients. This model preserves the appropriate properties of 2D-GARCH for modeling the curvelet coefficients while extends the dynamic formulation of 2D-GARCH model. Consequently, we design a maximum a-posteriori (MAP) estimator for estimating the clean image curvelet coefficients. Finally, we compare our proposed method with other denoising methods, and quantify the achieved performance improvement.