Relaxation Matching for Georegistration of Aerial and Satellite Imagery

In this paper, we present a novel approach for image-to-map registration through graph matching. In our graph-based matching approach road networks are presented as graphs, and registration is performed through the optimal mapping between two such graphs. A global compatibility function is formulated to measure the overall goodness of correspondence between two graphs, and is optimized through the use of continuous relaxation labeling. The main advantages of our approach lie on its invariance to translation, rotation and scale differences (through the use of appropriate attributes), as well as on its use of network structure to reduce the ambiguity in search space for inexact matching. Furthermore, our approach requires no user-defined threshold to justify local matches. In this paper we present the theoretical background behind our approach, and experimental results to demonstrate its performance.