A new SAR image change detection algorithm based on texture feature

Due to its all-weather all-time characteristics, detecting changes in land cover through time using synthetic aperture radar (SAR) images is one of the most appealing applications of remote sensing technology. And there are many SAR image change detection algorithms have been proposed. Among them, the methods based on texture feature have received more and more attention, because they can depress the effect of SAR image speckle noise. In this paper, we present a new texture feature change detection algorithm for SAR image. With the help of direction filter bank, we first decompose each original SAR image into four direction subband images. Then, we extract the texture feature by computing fractal dimension of each pixel neighborhood in each subband images. Finally, the change area is detected by comparing texture features of two times SAR images. Experimental results reveal that the proposed approach can get better performance than the widely used gray-level coocurrence matrix (GLCM) method.