An　improved　algorithm　based　on　wavelet　thresholding　is　proposed　for　suppressing　speckle　noise　of　synthetic　aperture　radar　image.　The　method　uses　the　local　statistics　of　multiscale　wavelet　coefficients　to　obtain　estimated　parameters　and　thresholds　based　on　BayesShrink　threshold.　Moreover,　it　employs　the　method　of　enhanced　Lee　filter　to　smooth　the　homogeneous　areas　and　preserve　the　point-target　areas.　Results　demonstrate　that　compared　with　the　results　of　enhanced　Lee　filter,　traditional　soft-thresholding　of　discrete　wavelet　transform　(DWT)　and　BayesShrink　soft-thresholding,　the　improved　algorithm　provides　the　largest　equivalent　number　of　looks　(ENL)　and　edge　sustain　index　(ESI),　which　indicates　it　can　not　only　reduce　the　speckle　but　also　preserve　edges　and　details.