Abstract: To avoid needing a preprocessing for reducing the speckle noise, and meanwhile overcome the limitation of selecting the distribution model, we integrate the characteristics of the DI with an interactive segmentation method not referring to any distribution assumption to generate the change detection maps corresponding to the different “seeds”, and then use a voting competition strategy to fuse those results to generate the final change detection map. During the segmenting, the feature of each pixel is set as a vector consisted of the corresponding intensities in the DI and each scale representation of the DI given by the use of stationary wavelet transform (SWT). This kind of features and the decision level fusion make our proposed method a certain robust to the speckle noise. The results tested on real SAR datasets and obtained under the situation that there is no despeckling preprocessing to SAR images, those performances are better than that of other techniques.