Abstract: An improved maximum a posteriori (MAP) estimator based on the heavy-tailed alpha-stable distribution is proposed. This method performs better in both speckle reduction and feature preservation, since the alpha-distribution model is used and a coefficient depending on the variance of the noise is introduced. Besides, the stationary wavelet transform (SWT) is used to avoid the artifacts as a result of Gibbs phenomenon. Finally, the proposed method is compared with the current state-of-the-art soft thresholding techniques and other local statistics filters applied on real SAR image and the proposed method is confirmed to be more adaptive, and it performs more effective in terms of both quantitive parameters and reconstructive image.

Key words: SAR image despeckling, wavelet analysis, symmetric alpha-stable distribution, MAP estimation