A keypoint approach for change detection between SAR images based on graph theory

This paper investigates the problem of change detection in multitemporal Synthetic Aperture Radar (SAR) images. Our proposition is to perform a keypoint-based algorithm to detect land-cover changes between two SAR images employing the graph theory combined with the log-ratio operator. First, a set of feature points is extracted from one of the two images. A weighted graph is then constructed to connect these keypoints based on their similarity measures from this first image. Based on this graph, our motivation is to measure the coherence between the information carried by the two images. In other words, the change level will depend on how much the second image still conforms to the graph constructed from the first image. Furthermore, due to the presence of speckle noise, the log-ratio operator will be exploited to replace the image difference operator. Experiments performed on real SAR images using the proposed algorithm provide very promising and competitive results compared to classical methods.