Wireless network virtualization and self-backhauled small-cell networks are considered as two promising technologies to enable next generation wireless networks. In this paper, we propose an integrated framework for the wireless network virtualization problem under the frequency division duplexing self-backhaul mechanism. In this context, the dynamic virtual resource allocation issue is formulated as an optimization problem, aiming at the maximization of the average total utility of the virtualized network. In the mean time, the network queue stability, the minimum average data rate for each service provider and the capacity constraint of the backhaul link are expected to be satisfied. To solve this optimization problem, the Lyapunov optimization method is adopted to overcome the challenge from the stochastic properties exhibited in the original problem, and a real-time scheduling algorithm is then investigated based on the current queue status information and the channel state information. Specifically, a Lagrange dual decomposition and an improved particle swarm optimization with random mutation are employed to ensure the effectiveness of the proposed algorithm. Simulation results are presented to verify the proposed algorithm.