Example-based Road Network Synthesis

We present a novel method for automatically synthesizing road networks that are perceptually similar to given example road networks. Our algorithm can grow a new road network from an initial node or expand an existing network. For an unfinished node in the network, we search a set of candidate nodes in the example road networks that have similar node topology, and then find a best one by neighborhood edge matching. Rather than simply copy edges incident to the best matched node, we consider the local constraints including obstacle avoidance and relationships to existing roads. Our results demonstrate that the method produces good results for various typical street patterns. Furthermore, we can blend styles of road networks in an intuitive and easy to control manner.