SAR images are subject to speckle effects which considerably reduce the detection of useful details within the scenes. Based on the analysis of the local speckle statistic characteristics of SAR image （such as relative standard deviation C u ） and algorithms of spatial filtering

an adaptive speckle filtering method is presented in this paper. Firstly

the special pixel neighborhood model is used in the selection of filtering windows

with which the filtering window is divided into 8 mutually exclusive sub windows and the central pixel is the only repetitions pixel in each sub window. Then

for every window to be filtered

the local relative standard deviation C i is calculated in all sub windows and used as the criteria to select the sub windows with C i

<

C u . Those selected sub windows are merged as a filtering area. Finally

recalculating relative standard deviation C x within the filtering area and different filter was used according to the difference of C x . In none of sub windows was selected or C x is larger than a threshold

the size of window should be reduced. If the size of window is 3 by 3 pixels

the central pixel should be replaced with the average of the nearest four pixels around it. The method was applied to both single

l

ook and multi look ERS 1/2 SAR images. The results show that the presented approach has the advantages both in speckle filtering and the preservation of image details.