As the wireless network scales up in size and complexity, the need to study the scalability and behaviors of these networks and their protocols becomes essential. Opportunistic routing utilizes broadcast nature of wireless network, and significantly increases the unicast throughput. However, all of the current opportunistic routing protocols have to rely on the whole topology information. This indeed restricts to applying the opportunistic routing to large-scale wireless networks, due to the huge cost of the control overheads needed for building a network graph at each node. In this paper, we propose the localized opportunistic routing (LOR) protocol, which utilizes the distributed minimum transmission selection (MTS) algorithm to partition the topology into several nested close-node-sets (CNS) with local information. It can locally realize the optimal opportunistic routing for large-scale wireless networks with low control overhead cost. Extensive simulation results show that in large-scale wireless environments, LOR can dramatically improve the performances over ExOR, MORE, in terms of control overhead, end-to-end delay and throughputs.