The success of vehicular networks is highly dependent on the coverage of messages, which refers to the trajectory of messages over time. Many of the existing works primarily performed in 1-D environments (i.e., straight highways) and/or merely focused on vehicle-to-vehicle (V2V) communications to enhance the coverage in a given road network. Consequently, there still lacks a clear comprehension of using road infrastructures (e.g., roadside units), to improve the message coverage in 2-D environments (i.e., urban areas). In this paper, we study the problem of optimal utilization of roadside units in 2-D environments. Specifically, we develop a message coverage maximization algorithm (MCMA) that carefully deploys the roadside units to achieve the maximum message coverage. Considering that the vehicle density is heterogeneous across the road networks, we study the message coverage for V2V networks by deriving analytical lower bounds of message dissemination distance for areas with different vehicle densities. The MCMA then utilizes the derived lower bounds to estimate the minimum spacing allowed between neighbor roadside units based on the prevailing traffic stream and delay constraint of applications. In addition, we propose a disseminator selection algorithm for infrastructure-based urban vehicular networks to further improve the message coverage. By selecting desired types of applications (i.e., safety and non-safety), we obtain two different roadside unit deployment sets from MCMA for the evaluation purpose. Extensive simulation studies show that MCMA outperforms the alternative algorithms in terms of the message coverage and message dissemination speed. The results also demonstrate that MCMA improves traffic efficiency in a post-crash scenario.