Practical Considerations in Unsupervised Change Detection Using SAR Images

Change detection has a wide range of applications in hurricane damage assessment, urban growth monitoring, etc. It is well-known that synthetic aperture radar (SAR) can penetrate clouds and can work well under all weather conditions. However, SAR images also contain a lot of speckle noise that seriously affects change detection performance. In this paper, we focus on change detection using SAR images. In particular, we propose a new change detection algorithm that is applicable to both optical and SAR images. The detection performance of the algorithm was compared with a number of existing algorithms in the literature. Preliminary results using actual SAR images are encouraging. Most importantly, it was observed that effective change detection using SAR images require good denoising and post-processing algorithms in order to achieve decent performance.