Abstract: This paper mainly deals with a new speckle reduction algorithm for high resolution SAR images based on wavelet transform. First of all, the mechanism of generating speckle in SAR image is explained. Then why a great deal of edge and texture were damaged when the speckle of high resolution SAR images were reduced by some traditional methods is discussed, and the advantage and the necessity of speckle reduction based on wavelet transform for reducing speckle are shown ; Secondly, it is expounded in detail how to select the size of a filter window basing on local statistics in each subband and the scale of decomposition for reducing speckle noise, in wavelet transform domain; Finally, Experimental results show that proposed method has better reducing speckle and edge preservation performance over the conventional filter with a window of fixed size.