Abstract: The paper has firstly discussed how a right wavelet base can be chosen，and decomposes the SAR image and optical image using the lifting scheme wavelet.Then，the low-frequency images are integrated with the weighted average ways after the SAR low-frequency image is filtered using the regional average means.In order to restrain the speckle noise，the paper focuses on the high-frequency images fusion，and presents a kind of the fusion methods that the windows can be changed according to the noise features，which can make the targets detection and recognition well.Finally，the fusion image is used to extract the features.The result shows that the fusion image is better than the original SAR image for the feature extraction.